Megatest

Changes On Branch e4bc65a729a9df54
Login

Changes In Branch v1.70-captain-ulex Through [e4bc65a729] Excluding Merge-Ins

This is equivalent to a diff from 269f41c0b0 to e4bc65a729

2020-01-04
16:45
Pulled in compilation fixes from refactor02. check-in: 337a4b27f1 user: matt tags: v1.70-captain-ulex, v1.70-defunct-try
16:36
Initial sysmon stuff check-in: e4bc65a729 user: matt tags: v1.70-captain-ulex, v1.70-defunct-try
2020-01-02
21:39
wip check-in: b95526b7f3 user: matt tags: v1.70-captain-ulex, v1.70-defunct-try
09:36
Fixed imports so dashboard launches Leaf check-in: 850872189d user: jmoon18 tags: v1.70-refactor02, v1.70-defunct-try
2019-12-31
23:20
Beginnings of captain based ulex check-in: 3dae400136 user: matt tags: v1.70-captain-ulex, v1.70-defunct-try
16:19
Added runsmod to eval-string in megatest.scm check-in: 269f41c0b0 user: mrwellan tags: v1.70-refactor02, v1.70-defunct-try
15:05
Fixed imports for dashboard check-in: 3bb6e533d4 user: jmoon18 tags: v1.70-refactor02, v1.70-defunct-try

Modified commonmod.scm from [838137cf51] to [106c49900b].

1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
    ;; TEST DATA
    test-data-rollup
    csv->test-data

    ;; MISC
    sync-inmem->db

    ;; TESTMETA
    testmeta-add-record
    testmeta-update-field

    ;; TASKS
    tasks-add
    tasks-set-state-given-param-key
    ))

;;======================================================================
;; ALLDATA
;;======================================================================
;;
;; attempt to consolidate a bunch of global information into one struct to toss around
(defstruct alldat
  ;; misc
  (denoise           (make-hash-table))
  (areapath          #f) ;; i.e. toppath
  (mtconfig          #f)
  (log-port          #f)
  (areadat           #f) ;; i.e. runremote
  (rmt-mutex         (make-mutex))
  (db-sync-mutex     (make-mutex))
  (db-with-db-mutex  (make-mutex))
  (read-only-queries api:read-only-queries)
  (write-queries     api:write-queries)
  (max-api-process-requests 0)
  (api-process-request-count 0)







|











|







|







1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
    ;; TEST DATA
    test-data-rollup
    csv->test-data

    ;; MISC
    sync-inmem->db

    ;; TESTMETAl
    testmeta-add-record
    testmeta-update-field

    ;; TASKS
    tasks-add
    tasks-set-state-given-param-key
    ))

;;======================================================================
;; ALLDATA
;;======================================================================

;; attempt to consolidate a bunch of global information into one struct to toss around
(defstruct alldat
  ;; misc
  (denoise           (make-hash-table))
  (areapath          #f) ;; i.e. toppath
  (mtconfig          #f)
  (log-port          #f)
  (ulexdat           #f) ;; connection to the databases
  (rmt-mutex         (make-mutex))
  (db-sync-mutex     (make-mutex))
  (db-with-db-mutex  (make-mutex))
  (read-only-queries api:read-only-queries)
  (write-queries     api:write-queries)
  (max-api-process-requests 0)
  (api-process-request-count 0)

Modified dashboard.scm from [bb015bee51] to [13566ec9e6].

53
54
55
56
57
58
59

60
61
62
63
64
65
66
67
68
69

70
71
72
73
74
75
76
(declare (uses ulex.import))
(declare (uses cookie.import))
(declare (uses stml2.import))
(declare (uses ducttape-lib.import))
(declare (uses mtconfigf.import))

(declare (uses gutilsmod))

(declare (uses megamod))
(declare (uses commonmod))
(declare (uses rmtmod))
(declare (uses runsmod))
(declare (uses dbmod))
(declare (uses testsmod))
(declare (uses dcommonmod))


(import gutilsmod)

(import megamod)
(import commonmod)
(import rmtmod)
(import runsmod)
(import dbmod)
(import testsmod)
(import dcommonmod)







>










>







53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
(declare (uses ulex.import))
(declare (uses cookie.import))
(declare (uses stml2.import))
(declare (uses ducttape-lib.import))
(declare (uses mtconfigf.import))

(declare (uses gutilsmod))
(declare (uses servermod))
(declare (uses megamod))
(declare (uses commonmod))
(declare (uses rmtmod))
(declare (uses runsmod))
(declare (uses dbmod))
(declare (uses testsmod))
(declare (uses dcommonmod))


(import gutilsmod)
(import servermod)
(import megamod)
(import commonmod)
(import rmtmod)
(import runsmod)
(import dbmod)
(import testsmod)
(import dcommonmod)
1416
1417
1418
1419
1420
1421
1422




























































































1423
1424
1425
1426
1427
1428
1429
       (iup:hbox
        (iup:button "Hide All" #:action (lambda (obj)
                                          (for-each (lambda (graph-cell)
                                                      (let* ((graph-dat   (hash-table-ref (dboard:tabdat-graph-cell-table tabdat) graph-cell)))
                                                        (dboard:graph-dat-flag-set! graph-dat #f)))
                                                    (hash-table-keys (dboard:tabdat-graph-cell-table tabdat)))))))
      ))))





























































































;;======================================================================
;; R U N
;;======================================================================
;;
;; display and manage a single run at a time








>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
       (iup:hbox
        (iup:button "Hide All" #:action (lambda (obj)
                                          (for-each (lambda (graph-cell)
                                                      (let* ((graph-dat   (hash-table-ref (dboard:tabdat-graph-cell-table tabdat) graph-cell)))
                                                        (dboard:graph-dat-flag-set! graph-dat #f)))
                                                    (hash-table-keys (dboard:tabdat-graph-cell-table tabdat)))))))
      ))))

;;======================================================================
;; SYSTEM MONITOR - Track relevant Megatest processes
;;======================================================================
;;
;; A gui for launching tests
;;
(define (dashboard:sysmon commondat tabdat #!key (tab-num #f))
  (let* ((drawing               (vg:drawing-new))
	 (sysmon-tab-updater (lambda ()	
			       (debug:catch-and-dump 
				(lambda ()
				  (let ((tabdat        (dboard:common-get-tabdat commondat tab-num: tab-num)))
				    (if tabdat
					(let ((last-data-update (dboard:tabdat-last-data-update tabdat))
					      (now-time         (current-seconds)))
					  (dashboard:sysmon-canvas-updater commondat tabdat tab-num) ;; redraws canvas on user input (I think...)
					  ))))
				   "dashboard:sysmon-tab-updater")))
	 (key-listboxes #f) ;; 
	 (update-keyvals (lambda ()
			   (dboard:target-updater tabdat))))
    (dboard:tabdat-drawing-set! tabdat drawing)
    (dboard:commondat-add-updater commondat run-times-tab-updater tab-num: tab-num)
    (iup:split
     #:orientation "VERTICAL" ;; "HORIZONTAL"
     #:value 150
     (iup:vbox
      (iup:split
       #:orientation "HORIZONTAL"
       #:value 800
      (let* ((cnv-obj (iup:canvas 
		       ;; #:size "250x250" ;; "500x400"
		       #:expand "YES"
		       #:scrollbar "YES"
		       #:posx "0.5"
		       #:posy "0.5"
		       #:action (make-canvas-action
				  (lambda (c xadj yadj)
				    (debug:catch-and-dump
				     (lambda ()
				       (if (not (dboard:tabdat-cnv tabdat))
					   (let ((cnv     (dboard:tabdat-cnv tabdat)))
					     (dboard:tabdat-cnv-set! tabdat c)
					     (vg:drawing-cnv-set! (dboard:tabdat-drawing tabdat)
								  (dboard:tabdat-cnv tabdat))))
				       (let ((drawing (dboard:tabdat-drawing tabdat))
					     (old-xadj (dboard:tabdat-xadj   tabdat))
					     (old-yadj (dboard:tabdat-yadj   tabdat)))
					 (if (not (and (eq? xadj old-xadj)(eq? yadj old-yadj)))
					     (begin
					       ;; (print  "xadj: " xadj " yadj: " yadj "changed: "(eq? xadj old-xadj) " " (eq? yadj old-yadj))
					       (dboard:tabdat-view-changed-set! tabdat #t)
					       (dboard:tabdat-xadj-set! tabdat (* -2000 (- xadj 0.5)))
					       (dboard:tabdat-yadj-set! tabdat (*  2000 (- yadj 0.5)))
					       ))))
				     "iup:canvas action")))
		       #:wheel-cb  (lambda (obj step x y dir) ;; dir is 4 for up and 5 for down. I think.
				     (debug:catch-and-dump
				      (lambda ()
					(let* ((drawing (dboard:tabdat-drawing tabdat))
					       (scalex  (vg:drawing-scalex drawing)))
					  (dboard:tabdat-view-changed-set! tabdat #t)
					  ;; (print "step: " step " x: " x " y: " y " dir: " dir " scalex: " scalex)
					  (vg:drawing-scalex-set! drawing
								  (+ scalex
								     (if (> step 0)
									 (* scalex  0.02)
									 (* scalex -0.02))))))
				      "wheel-cb"))
		       )))
	cnv-obj)
      )))))

;; run times canvas updater
;;
(define (dashboard:sysmon-canvas-updater commondat tabdat tab-num)
  (let ((cnv (dboard:tabdat-cnv tabdat))
	(dwg (dboard:tabdat-drawing tabdat))
	(mtx (dboard:tabdat-runs-mutex tabdat))
	(vch (dboard:tabdat-view-changed tabdat)))
    (if (and cnv dwg vch)
	(begin
	  (vg:drawing-xoff-set! dwg (dboard:tabdat-xadj tabdat))
	  (vg:drawing-yoff-set! dwg (dboard:tabdat-yadj tabdat))
	  (mutex-lock! mtx)
	  (canvas-clear! cnv)
	  (vg:draw dwg tabdat)
	  (mutex-unlock! mtx)
	  (dboard:tabdat-view-changed-set! tabdat #f)))))
  


;;======================================================================
;; R U N
;;======================================================================
;;
;; display and manage a single run at a time

2188
2189
2190
2191
2192
2193
2194

2195
2196
2197
2198
2199
2200
2201

(define (make-dashboard-buttons commondat) ;;  runs-sum-dat new-view-dat)
  (let* ((stats-dat       (dboard:tabdat-make-data))
	 (runs-dat        (dboard:tabdat-make-data))
	 (onerun-dat      (dboard:tabdat-make-data)) ;; name for run-summary structure 
	 (runcontrols-dat (dboard:tabdat-make-data))
	 (runtimes-dat    (dboard:tabdat-make-data))

	 (nruns           (dboard:tabdat-numruns runs-dat))
	 (ntests          (dboard:tabdat-num-tests runs-dat))
	 (keynames        (dboard:tabdat-dbkeys runs-dat))
	 (nkeys           (length keynames))
	 (runsvec         (make-vector nruns))
	 (header          (make-vector nruns))
	 (lftcol          (make-vector ntests))







>







2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296

(define (make-dashboard-buttons commondat) ;;  runs-sum-dat new-view-dat)
  (let* ((stats-dat       (dboard:tabdat-make-data))
	 (runs-dat        (dboard:tabdat-make-data))
	 (onerun-dat      (dboard:tabdat-make-data)) ;; name for run-summary structure 
	 (runcontrols-dat (dboard:tabdat-make-data))
	 (runtimes-dat    (dboard:tabdat-make-data))
	 (sysmon-dat      (dboard:tabdat-make-data))
	 (nruns           (dboard:tabdat-numruns runs-dat))
	 (ntests          (dboard:tabdat-num-tests runs-dat))
	 (keynames        (dboard:tabdat-dbkeys runs-dat))
	 (nkeys           (length keynames))
	 (runsvec         (make-vector nruns))
	 (header          (make-vector nruns))
	 (lftcol          (make-vector ntests))
2406
2407
2408
2409
2410
2411
2412

2413
2414
2415
2416
2417
2418
2419
2420

2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438

2439
2440
2441
2442
2443
2444
2445
			  (dashboard:summary commondat stats-dat tab-num: 0)
			  runs-view
			  (dashboard:runs-summary commondat onerun-dat tab-num: 2)
			  ;; (dashboard:new-view db data new-view-dat tab-num: 3)
			  (dashboard:run-controls commondat runcontrols-dat tab-num: 3)
			  (dashboard:run-times commondat runtimes-dat tab-num: 4)
			  ;; ;; (dashboard:runs-summary commondat onerun-dat tab-num: 4)

			  additional-views
			  )))
	;; (set! (iup:callback tabs tabchange-cb:) (lambda (a b c)(print "SWITCHED TO TAB: " a " " b " " c)))
	(iup:attribute-set! tabs "TABTITLE0" "Summary")
	(iup:attribute-set! tabs "TABTITLE1" "Runs")
	(iup:attribute-set! tabs "TABTITLE2" "Run Summary")
	(iup:attribute-set! tabs "TABTITLE3" "Run Control")
	(iup:attribute-set! tabs "TABTITLE4" "Run Times")

	;; (iup:attribute-set! tabs "TABTITLE3" "New View")
	;; (iup:attribute-set! tabs "TABTITLE4" "Run Control")

	;; set the tab names for user added tabs
	(for-each
	 (lambda (tab-info)
	   (iup:attribute-set! tabs (conc "TABTITLE" (car tab-info)) (cdr tab-info)))
	 additional-tabnames)
	
	(iup:attribute-set! tabs "BGCOLOR" "190 190 190")
	;; make the iup tabs object available (for changing color for example)
	(dboard:commondat-hide-not-hide-tabs-set! commondat tabs)
	;; now set up the tabdat lookup
	(dboard:common-set-tabdat! commondat 0 stats-dat)
	(dboard:common-set-tabdat! commondat 1 runs-dat)
	(dboard:common-set-tabdat! commondat 2 onerun-dat)
	(dboard:common-set-tabdat! commondat 3 runcontrols-dat)
	(dboard:common-set-tabdat! commondat 4 runtimes-dat)


	(iup:vbox
	 tabs
	 ;; controls
	 ))))
    (vector keycol lftcol header runsvec)))








>








>


















>







2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
			  (dashboard:summary commondat stats-dat tab-num: 0)
			  runs-view
			  (dashboard:runs-summary commondat onerun-dat tab-num: 2)
			  ;; (dashboard:new-view db data new-view-dat tab-num: 3)
			  (dashboard:run-controls commondat runcontrols-dat tab-num: 3)
			  (dashboard:run-times commondat runtimes-dat tab-num: 4)
			  ;; ;; (dashboard:runs-summary commondat onerun-dat tab-num: 4)
			  (dashboard:sysmon commondat sysmon-dat tab-num: 5)
			  additional-views
			  )))
	;; (set! (iup:callback tabs tabchange-cb:) (lambda (a b c)(print "SWITCHED TO TAB: " a " " b " " c)))
	(iup:attribute-set! tabs "TABTITLE0" "Summary")
	(iup:attribute-set! tabs "TABTITLE1" "Runs")
	(iup:attribute-set! tabs "TABTITLE2" "Run Summary")
	(iup:attribute-set! tabs "TABTITLE3" "Run Control")
	(iup:attribute-set! tabs "TABTITLE4" "Run Times")
	(iup:attribute-set! tabs "TABTITLE5" "Sysmon")
	;; (iup:attribute-set! tabs "TABTITLE3" "New View")
	;; (iup:attribute-set! tabs "TABTITLE4" "Run Control")

	;; set the tab names for user added tabs
	(for-each
	 (lambda (tab-info)
	   (iup:attribute-set! tabs (conc "TABTITLE" (car tab-info)) (cdr tab-info)))
	 additional-tabnames)
	
	(iup:attribute-set! tabs "BGCOLOR" "190 190 190")
	;; make the iup tabs object available (for changing color for example)
	(dboard:commondat-hide-not-hide-tabs-set! commondat tabs)
	;; now set up the tabdat lookup
	(dboard:common-set-tabdat! commondat 0 stats-dat)
	(dboard:common-set-tabdat! commondat 1 runs-dat)
	(dboard:common-set-tabdat! commondat 2 onerun-dat)
	(dboard:common-set-tabdat! commondat 3 runcontrols-dat)
	(dboard:common-set-tabdat! commondat 4 runtimes-dat)
	(dboard:common-set-tabdat! commondat 5 sysmon-dat)

	(iup:vbox
	 tabs
	 ;; controls
	 ))))
    (vector keycol lftcol header runsvec)))

Modified rmtmod.scm from [90119f80f9] to [5c3bd41b91].

35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
(import (prefix ulex ulex:))

(import commonmod)
(import itemsmod)
(import apimod)
(import dbmod)

;; (include "rmt-inc.scm")

;;
;; THESE ARE ALL CALLED ON THE CLIENT SIDE!!!
;;

;; generate entries for ~/.megatestrc with the following
;;
;;  grep define ../rmt.scm | grep rmt: |perl -pi -e 's/\(define\s+\((\S+)\W.*$/\1/'|sort -u







<
<







35
36
37
38
39
40
41


42
43
44
45
46
47
48
(import (prefix ulex ulex:))

(import commonmod)
(import itemsmod)
(import apimod)
(import dbmod)



;;
;; THESE ARE ALL CALLED ON THE CLIENT SIDE!!!
;;

;; generate entries for ~/.megatestrc with the following
;;
;;  grep define ../rmt.scm | grep rmt: |perl -pi -e 's/\(define\s+\((\S+)\W.*$/\1/'|sort -u
70
71
72
73
74
75
76





77








78



79







80






81
82
83
84
85
86
87
			#f)))
	  (if cinfo
	      cinfo
	      (if (server:check-if-running areapath)
		  (client:setup areapath)
		  #f))))






(define *send-receive-mutex* (make-mutex)) ;; should have separate mutex per run-id












(define (rmt:send-receive cmd rid params #!key (attemptnum 1)(area-dat #f)) ;; start attemptnum at 1 so the modulo below works as expected







  (rmt:open-qry-close-locally cmd 0 params))






;;   
;; ;; #;(common:telemetry-log (conc "rmt:"(->string cmd))
;; ;; #;(define (rmt:send-receive cmd rid params #!key (attemptnum 1)(area-dat #f)) ;; start attemptnum at 1 so the modulo below works as expected
;; ;; 
;; ;; #;(common:telemetry-log (conc "rmt:"(->string cmd))
;; ;; payload: `((rid . ,rid)
;; ;; (params . ,params)))







>
>
>
>
>
|
>
>
>
>
>
>
>
>

>
>
>

>
>
>
>
>
>
>
|
>
>
>
>
>
>







68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
			#f)))
	  (if cinfo
	      cinfo
	      (if (server:check-if-running areapath)
		  (client:setup areapath)
		  #f))))

;; return the handle struct for sending queries to a specific database
;;  - initializes the connection object if this is the first access
;;    - finds the "captain" and asks who to talk to for the given dbfname
;;    - establishes the connection to the current dbowner
;;
(define (rmt:connect alldat dbfname)
  (let* ((ulexdat    (let ((uconn (alldat-ulexdat alldat)))
		       (if uconn
			   uconn
	  		   (let* ((new-ulexdat (ulex:setup))) ;; establish connection to ulex
			     (alldat-ulexdat-set! alldat new-ulexdat)
			     (rmt:setup-ulex alldat)
			     new-ulexdat)))))
    (ulex:connect ulexdat dbfname)))

;; set up a connection to the current owner of the dbfile associated with rid
;; then send the query to that dbfile owner and wait for a response.
;;
(define (rmt:send-receive cmd rid params #!key (attemptnum 1)(area-dat #f)) ;; start attemptnum at 1 so the modulo below works as expected
  (let* ((alldat   *alldat*)
	 (areapath (alldat-areapath alldat))
	 (dbfname  (if (or (not rid)(< rid 1)) ;; this is the criteria for "main.db"
		       "main.db"
		       (conc rid ".db")))
	 (dbfile   (conc areapath "/.db/" dbfname))
	 (ulexconn (rmt:connect alldat dbfname)))  
    (rmt:open-qry-close-locally cmd 0 params)))

;; setup the remote calls
(define (rmt:setup-ulex alldat)
  (let ((udata (alldat-ulexdat alldat)))
    (ulex:register-handler udata 'ping common:get-full-version)
  ))
;;   
;; ;; #;(common:telemetry-log (conc "rmt:"(->string cmd))
;; ;; #;(define (rmt:send-receive cmd rid params #!key (attemptnum 1)(area-dat #f)) ;; start attemptnum at 1 so the modulo below works as expected
;; ;; 
;; ;; #;(common:telemetry-log (conc "rmt:"(->string cmd))
;; ;; payload: `((rid . ,rid)
;; ;; (params . ,params)))

Modified ulex/ulex.scm from [ef093072a2] to [63bf7783f2].

22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68

69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
;;   See README in the distribution at https://www.kiatoa.com/fossils/ulex
;; NOTES:
;;   Why sql-de-lite and not say, dbi?  - performance mostly, then simplicity.
;;
;;======================================================================

;; (use rpc pkts mailbox sqlite3)
  
(module ulex
    *
     ;;    #;(
     ;;     ;; areas
     ;;     make-area
     ;;     area->alist
     ;;     ;; server
     ;;     launch
     ;;     update-known-servers
     ;;     shutdown
     ;;     get-best-server
     ;;     ;; client side
     ;;     call
     ;;     ;; queries, procs and system commands (i.e. workers)
     ;;     register
     ;;     register-batch
     ;;     ;; database - note: most database stuff need not be exposed, these calls may be removed from exports in future
     ;;     open-db
     ;;     ;; ports
     ;;     pl-find-port
     ;;     pl-get-prev-used-port
     ;;     pl-open-db
     ;;     pl-open-run-close
     ;;     pl-release-port
     ;;     pl-set-port
     ;;     pl-take-port
     ;;     pl-is-port-available
     ;;     pl-get-port-state
     ;;     ;; system
     ;;     get-normalized-cpu-load
	 
     ;; )

(import scheme posix chicken data-structures ports extras files mailbox)
(import rpc srfi-18 pkts matchable regex
	typed-records srfi-69 srfi-1
	srfi-4 regex-case
	(prefix sqlite3 sqlite3:)
	foreign

	tcp) ;; ulex-netutil)

;;======================================================================
;; D E B U G   H E L P E R S
;;======================================================================
    
(define (dbg> . args)
  (with-output-to-port (current-error-port)
    (lambda ()
      (apply print "dbg> " args))))

(define (debug-pp . args)
  (if (get-environment-variable "ULEX_DEBUG")
      (with-output-to-port (current-error-port)
	(lambda ()
	  (apply pp args)))))

(define *default-debug-port* (current-error-port))

(define (sdbg> fn stage-name stage-start stage-end start-time . message)
  (if (get-environment-variable "ULEX_DEBUG")
      (with-output-to-port *default-debug-port* 
	(lambda ()
	  (apply print "ulex:" fn " " stage-name " took " (- (if stage-end stage-end (current-milliseconds)) stage-start) " ms. "
		 (if start-time
		     (conc "total time " (- (current-milliseconds) start-time)
			   " ms.")
		     "")
		 message
		 )))))

;;======================================================================
;; M A C R O S
;;======================================================================
;; iup callbacks are not dumping the stack, this is a work-around
;;

;; Some of these routines use:
;;
;;     http://www.cs.toronto.edu/~gfb/scheme/simple-macros.html
;;
;; Syntax for defining macros in a simple style similar to function definiton,
;;  when there is a single pattern for the argument list and there are no keywords.
;;
;; (define-simple-syntax (name arg ...) body ...)
;;

(define-syntax define-simple-syntax
  (syntax-rules ()
    ((_ (name arg ...) body ...)
     (define-syntax name (syntax-rules () ((name arg ...) (begin body ...)))))))

(define-simple-syntax (catch-and-dump proc procname)
  (handle-exceptions
   exn
   (begin
     (print-call-chain (current-error-port))
     (with-output-to-port (current-error-port)
       (lambda ()
         (print ((condition-property-accessor 'exn 'message) exn))
         (print "Callback error in " procname)
         (print "Full condition info:\n" (condition->list exn)))))
   (proc)))


;;======================================================================
;;  R E C O R D S
;;======================================================================

;; information about me as a server
;;
(defstruct area
  ;; about this area
  (useportlogger #f)
  (lowport       32768)
  (server-type   'auto)  ;; auto=create up to five servers/pkts, main=create pkts, passive=no pkt (unless there are no pkts at all)
  (conn          #f)
  (port          #f)
  (myaddr        (get-my-best-address))
  pktid          ;; get pkt from hosts table if needed
  pktfile
  pktsdir
  dbdir
  (dbhandles     (make-hash-table)) ;; fname => list-of-dbh, NOTE: Should really never need more than one?
  (mutex         (make-mutex))
  (rtable        (make-hash-table)) ;; registration table of available actions
  (dbs           (make-hash-table)) ;; filename => random number, used for choosing what dbs I serve
  ;; about other servers
  (hosts         (make-hash-table)) ;; key => hostdat
  (hoststats     (make-hash-table)) ;; key => alist of fname => ( qcount . qtime )
  (reqs          (make-hash-table)) ;; uri => queue
  ;; work queues
  (wqueues       (make-hash-table)) ;; fname => qdat
  (stats         (make-hash-table)) ;; fname => totalqueries
  (last-srvup    (current-seconds)) ;; last time we updated the known servers
  (cookie2mbox   (make-hash-table)) ;; map cookie for outstanding request to mailbox of awaiting call
  (ready #f)
  (health        (make-hash-table)) ;; ipaddr:port => num failed pings since last good ping
  )

;; host stats
;;
(defstruct hostdat
  (pkt      #f)
  (dbload   (make-hash-table))  ;; "dbfile.db" => queries/min
  (hostload #f)                 ;; normalized load ( 5min load / numcpus )
  )

;; dbdat
;;
(defstruct dbdat
  (dbh    #f)
  (fname  #f)
  (write-access #f)
  (sths   (make-hash-table))  ;; hash mapping query strings to handles
  )

;; qdat
;;
(defstruct qdat
  (writeq  (make-queue))
  (readq   (make-queue))
  (rwq     (make-queue))
  (logq    (make-queue)) ;; do we need a queue for logging? yes, if we use sqlite3 db for logging
  (osshort (make-queue))
  (oslong  (make-queue))
  (misc    (make-queue)) ;; used for things like ping-full
  )

;; calldat
;;
(defstruct calldat
  (ctype 'dbwrite)
  (obj   #f)              ;; this would normally be an SQL statement e.g. SELECT, INSERT etc.
  (rtime (current-milliseconds)))

;; make it a global? Well, it is local to area module

(define *pktspec*
  `((server (hostname . h)
	    (port     . p)
	    (pid      . i)
	    (ipaddr   . a)
	    )
    (data   (hostname . h)  ;; sender hostname
	    (port     . p)  ;; sender port
	    (ipaddr   . a)  ;; sender ip
	    (hostkey  . k)  ;; sending host key - store info at server under this key
	    (servkey  . s)  ;; server key - this needs to match at server end or reject the msg
	    (format   . f)  ;; sb=serialized-base64, t=text, sx=sexpr, j=json
	    (data     . d)  ;; base64 encoded slln data
	    )))

;; work item
;;
(defstruct witem
  (rhost #f)   ;; return host
  (ripaddr #f) ;; return ipaddr
  (rport #f)   ;; return port
  (servkey #f) ;; the packet representing the client of this workitem, used by final send-message
  (rdat  #f)   ;; the request - usually an sql query, type is rdat
  (action #f)  ;; the action: immediate, dbwrite, dbread,oslong, osshort
  (cookie #f)  ;; cookie id for response
  (data   #f)  ;; the data payload, i.e. parameters
  (result #f)  ;; the result from processing the data
  (caller #f)) ;; the calling peer according to rpc itself

(define (trim-pktid pktid)
  (if (string? pktid)
      (substring pktid 0 4)
      "nopkt"))

(define (any->number num)
  (cond
   ((number? num) num)
   ((string? num) (string->number num))
   (else num)))

(use trace)
(trace-call-sites #t)

;;======================================================================
;; D A T A B A S E   H A N D L I N G 
;;======================================================================

;; look in dbhandles for a db, return it, else return #f
;;
(define (get-dbh acfg fname)
  (let ((dbh-lst (hash-table-ref/default (area-dbhandles acfg) fname '())))
    (if (null? dbh-lst)
	(begin
	  ;; (print "opening db for " fname)
	  (open-db acfg fname)) ;; Note that the handles get put back in the queue in the save-dbh calls
	(let ((rem-lst (cdr dbh-lst)))
	  ;; (print "re-using saved connection for " fname)
	  (hash-table-set! (area-dbhandles acfg) fname rem-lst)
	  (car dbh-lst)))))

(define (save-dbh acfg fname dbdat)
    ;; (print "saving dbh for " fname)
    (hash-table-set! (area-dbhandles acfg) fname (cons dbdat (hash-table-ref/default (area-dbhandles acfg) fname '()))))

;; open the database, if never before opened init it. put the handle in the
;; open db's hash table
;; returns: the dbdat
;;
(define (open-db acfg fname)
  (let* ((fullname     (conc (area-dbdir acfg) "/" fname))
	 (exists       (file-exists? fullname))
	 (write-access (if exists
			   (file-write-access? fullname)
			   (file-write-access? (area-dbdir acfg))))
	 (db           (sqlite3:open-database fullname))
	 (handler      (sqlite3:make-busy-timeout 136000))
	 )
    (sqlite3:set-busy-handler! db handler)
    (sqlite3:execute db "PRAGMA synchronous = 0;")
    (if (not exists) ;; need to init the db
	(if write-access
	    (let ((isql (get-rsql acfg 'dbinitsql))) ;; get the init sql statements
	      ;; (sqlite3:with-transaction
	      ;;  db
	      ;;  (lambda ()
		 (if isql
		     (for-each
		      (lambda (sql)
			(sqlite3:execute db sql))
		      isql)))
	    (print "ERROR: no write access to " (area-dbdir acfg))))
    (make-dbdat dbh: db fname: fname write-access: write-access)))

;; This is a low-level command to retrieve or to prepare, save and return a prepared statment
;; you must extract the db handle
;;
(define (get-sth db cache stmt)
  (if (hash-table-exists? cache stmt)
      (begin
	;; (print "Reusing cached stmt for " stmt)
	(hash-table-ref/default cache stmt #f))
      (let ((sth (sqlite3:prepare db stmt)))
	(hash-table-set! cache stmt sth)
	;; (print "prepared stmt for " stmt)
	sth)))

;; a little more expensive but does all the tedious deferencing - only use if you don't already
;; have dbdat and db sitting around
;;
(define (full-get-sth acfg fname stmt)
  (let* ((dbdat  (get-dbh acfg fname))
	 (db     (dbdat-dbh dbdat))
	 (sths   (dbdat-sths dbdat)))
    (get-sth db sths stmt)))

;; write to a db
;; acfg: area data
;; rdat: request data
;; hdat: (host . port)
;;
;; (define (dbwrite acfg rdat hdat data-in)
;;   (let* ((dbname (car data-in))
;; 	 (dbdat  (get-dbh acfg dbname))
;; 	 (db     (dbdat-dbh dbdat))
;; 	 (sths   (dbdat-sths dbdat))
;; 	 (stmt   (calldat-obj rdat))
;; 	 (sth    (get-sth db sths stmt))
;; 	 (data   (cdr data-in)))
;;     (print "dbname: " dbname " acfg: " acfg " rdat: " (calldat->alist rdat) " hdat: " hdat " data: " data)
;;     (print "dbdat: " (dbdat->alist dbdat))
;;     (apply sqlite3:execute sth data)
;;     (save-dbh acfg dbname dbdat)
;;     #t
;;     ))

(define (finalize-all-db-handles acfg)
  (let* ((dbhandles (area-dbhandles acfg))  ;; dbhandles is hash of fname ==> dbdat
	 (num       0))
    (for-each
     (lambda (area-name)
       (print "Closing handles for " area-name)
       (let ((dbdats (hash-table-ref/default dbhandles area-name '())))
	 (for-each
	  (lambda (dbdat)
	    ;; first close all statement handles
	    (for-each
	     (lambda (sth)
	       (sqlite3:finalize! sth)
	       (set! num (+ num 1)))
	     (hash-table-values (dbdat-sths dbdat)))
	    ;; now close the dbh
	    (set! num (+ num 1))
	    (sqlite3:finalize! (dbdat-dbh dbdat)))
	  dbdats)))
     (hash-table-keys dbhandles))
    (print "FINALIZED " num " dbhandles")))

;;======================================================================
;; W O R K   Q U E U E   H A N D L I N G 
;;======================================================================

(define (register-db-as-mine acfg dbname)
  (let ((ht (area-dbs acfg)))
    (if (not (hash-table-ref/default ht dbname #f))
	(hash-table-set! ht dbname (random 10000)))))
	
(define (work-queue-add acfg fname witem)
  (let* ((work-queue-start (current-milliseconds))
	 (action           (witem-action witem)) ;; NB the action is the index into the rdat actions
	 (qdat             (or (hash-table-ref/default (area-wqueues acfg) fname #f)
			       (let ((newqdat (make-qdat)))
				 (hash-table-set! (area-wqueues acfg) fname newqdat)
				 newqdat)))
	 (rdat             (hash-table-ref/default (area-rtable acfg) action #f)))
    (if rdat
	(queue-add!
	 (case (calldat-ctype rdat)
	   ((dbwrite)   (register-db-as-mine acfg fname)(qdat-writeq qdat))
	   ((dbread)    (register-db-as-mine acfg fname)(qdat-readq  qdat))
	   ((dbrw)      (register-db-as-mine acfg fname)(qdat-rwq    qdat))
	   ((oslong)    (qdat-oslong qdat))
	   ((osshort)   (qdat-osshort qdat))
	   ((full-ping) (qdat-misc  qdat))
	   (else
	    (print "ERROR: no queue for " action ". Adding to dbwrite queue.")
	    (qdat-writeq qdat)))
	 witem)
	(case action
	  ((full-ping)(qdat-misc qdat))
	  (else
	   (print "ERROR: No action " action " was registered"))))
    (sdbg> "work-queue-add" "queue-add" work-queue-start #f #f)
    #t)) ;; for now, simply return #t to indicate request got to the queue

(define (doqueue acfg q fname dbdat dbh)
  ;; (print "doqueue: " fname)
  (let* ((start-time (current-milliseconds))
	 (qlen       (queue-length q)))
    (if (> qlen 1)
	(print "Processing queue of length " qlen))
    (let loop ((count      0)
	       (responses '()))
      (let ((delta (- (current-milliseconds) start-time)))
	(if (or (queue-empty? q)
		(> delta 400)) ;; stop working on this queue after 400ms have passed
	    (list count delta responses) ;; return count, delta and responses list
	    (let* ((witem  (queue-remove! q))
		   (action (witem-action witem))
		   (rdat   (witem-rdat   witem))
		   (stmt   (calldat-obj rdat))
		   (sth    (full-get-sth acfg fname stmt))
		   (ctype  (calldat-ctype rdat))
		   (data   (witem-data   witem))
		   (cookie (witem-cookie witem)))
	      ;; do the processing and save the result in witem-result
	      (witem-result-set!
	       witem
	       (case ctype ;; action
		 ((noblockwrite) ;; blind write, no ack of success returned
		  (apply sqlite3:execute sth data)
		  (sqlite3:last-insert-rowid dbh))
		 ((dbwrite)      ;; blocking write   
		  (apply sqlite3:execute sth data)
		  #t)
		 ((dbread) ;; TODO: consider breaking this up and shipping in pieces for large query
		  (apply sqlite3:map-row (lambda x x) sth data))
		 ((full-ping)  'full-ping)
		 (else (print "Not ready for action " action) #f)))
	      (loop (add1 count)
		    (if cookie
			(cons witem responses)
			responses))))))))

;; do up to 400ms of processing on each queue
;; - the work-queue-processor will allow the max 1200ms of work to complete but it will flag as overloaded
;; 
(define (process-db-queries acfg fname)
  (if (hash-table-exists? (area-wqueues acfg) fname)
      (let* ((process-db-queries-start-time (current-milliseconds))
	     (qdat             (hash-table-ref/default (area-wqueues acfg) fname #f))
	     (queue-sym->queue (lambda (queue-sym)
				 (case queue-sym  ;; lookup the queue from qdat given a name (symbol)
				   ((wqueue)  (qdat-writeq qdat))
				   ((rqueue)  (qdat-readq  qdat))
				   ((rwqueue) (qdat-rwq    qdat))
				   ((misc)    (qdat-misc   qdat))
				   (else #f))))
	     (dbdat   (get-dbh acfg fname))
	     (dbh     (if (dbdat? dbdat)(dbdat-dbh dbdat) #f))
	     (nowtime (current-seconds)))
	;; handle the queues that require a transaction
	;;
	(map ;; 
	 (lambda (queue-sym)
	   ;; (print "processing queue " queue-sym)
	   (let* ((queue (queue-sym->queue queue-sym)))
	     (if (not (queue-empty? queue))
		 (let ((responses
			(sqlite3:with-transaction ;; todo - catch exceptions...
			 dbh
			 (lambda ()
			   (let* ((res (doqueue acfg queue fname dbdat dbh))) ;; this does the work!
			     ;; (print "res=" res)
			     (match res
			      ((count delta responses)
			       (update-stats acfg fname queue-sym delta count)
			       (sdbg> "process-db-queries" "sqlite3-transaction" process-db-queries-start-time #f #f)
			       responses) ;; return responses
			      (else
			       (print "ERROR: bad return data from doqueue " res)))
			     )))))
		   ;; having completed the transaction, send the responses.
		   ;; (print "INFO: sending " (length responses) " responses.")
		   (let loop ((responses-left responses))
		     (cond
		      ((null? responses-left)  #t)
		      (else
		       (let* ((witem    (car responses-left))
			      (response (cdr responses-left)))  
			 (call-deliver-response acfg (witem-ripaddr witem)(witem-rport witem)
						(witem-cookie witem)(witem-result witem)))
		       (loop (cdr responses-left))))))
		 )))
	 '(wqueue rwqueue rqueue))
	
	;; handle misc queue
	;;
	;; (print "processing misc queue")
	(let ((queue (queue-sym->queue 'misc)))
	  (doqueue acfg queue fname dbdat dbh))
	;; ....
	(save-dbh acfg fname dbdat)
	#t ;; just to let the tests know we got here
	)
      #f ;; nothing processed
      ))

;; run all queues in parallel per db but sequentially per queue for that db.
;;  - process the queues every 500 or so ms
;;  - allow for long running queries to continue but all other activities for that
;;    db will be blocked.
;;
(define (work-queue-processor acfg)
  (let* ((threads (make-hash-table))) ;; fname => thread
    (let loop ((fnames      (hash-table-keys (area-wqueues acfg)))
	       (target-time (+ (current-milliseconds) 50)))
      ;;(if (not (null? fnames))(print "Processing for these databases: " fnames))
      (for-each
       (lambda (fname)
	 ;; (print "processing for " fname)
	 ;;(process-db-queries acfg fname))
	 (let ((th (hash-table-ref/default threads fname #f)))
	   (if (and th (not (member (thread-state th) '(dead terminated))))
	       (begin
		 (print "WARNING: worker thread for " fname " is taking a long time.")
		 (print "Thread is in state " (thread-state th)))
	       (let ((th1 (make-thread (lambda ()
					 ;; (catch-and-dump
					 ;;  (lambda ()
					    ;; (print "Process queries for " fname)
					    (let ((start-time (current-milliseconds)))
					      (process-db-queries acfg fname)
					      ;; (thread-sleep! 0.01) ;; need the thread to take at least some time
					      (hash-table-delete! threads fname)) ;; no mutexes?
					    fname)
					  "th1"))) ;; ))
		 (hash-table-set! threads fname th1)
		 (thread-start! th1)))))
       fnames)
      ;; (thread-sleep! 0.1) ;; give the threads some time to process requests
      ;; burn time until 400ms is up
      (let ((now-time (current-milliseconds)))
	(if (< now-time target-time)
	    (let ((delta (- target-time now-time)))
	      (thread-sleep! (/ delta 1000)))))
      (loop (hash-table-keys (area-wqueues acfg))
	    (+ (current-milliseconds) 50)))))

;;======================================================================
;; S T A T S   G A T H E R I N G
;;======================================================================

(defstruct stat
  (qcount-avg  0)                  ;; coarse running average
  (qtime-avg   0)                  ;; coarse running average
  (qcount      0)                  ;; total
  (qtime       0)                  ;; total
  (last-qcount 0)                  ;; last 
  (last-qtime  0)                  ;; last
  (dbs        '())                 ;; list of db files handled by this node
  (when        0))                 ;; when the last query happened - seconds


(define (update-stats acfg fname bucket duration numqueries)
  (let* ((key   fname) ;; for now do not use bucket. Was: (conc fname "-" bucket)) ;; lazy but good enough
	 (stats (or (hash-table-ref/default (area-stats acfg) key #f)
		    (let ((newstats (make-stat)))
		      (hash-table-set! (area-stats acfg) key newstats)
		      newstats))))
    ;; when the last query happended (used to remove the fname from the active list)
    (stat-when-set! stats (current-seconds))
    ;; last values
    (stat-last-qcount-set! stats numqueries)
    (stat-last-qtime-set!  stats duration)
    ;; total over process lifetime
    (stat-qcount-set! stats (+ (stat-qcount stats) numqueries))
    (stat-qtime-set!  stats (+ (stat-qtime  stats) duration))
    ;; coarse average
    (stat-qcount-avg-set! stats (/ (+ (stat-qcount-avg stats) numqueries) 2))
    (stat-qtime-avg-set!  stats (/ (+ (stat-qtime-avg  stats) duration)   2))

    ;; here is where we add the stats for a given dbfile
    (if (not (member fname (stat-dbs stats)))
	(stat-dbs-set! stats (cons fname (stat-dbs stats))))

    ))

;;======================================================================
;; S E R V E R   S T U F F 
;;======================================================================

;; this does NOT return!
;;
(define (find-free-port-and-open acfg)
  (let ((port (or (area-port acfg) 3200)))
    (handle-exceptions
	exn
	(begin
	  (print "INFO: cannot bind to port " (rpc:default-server-port) ", trying next port")
	  (area-port-set! acfg (+ port 1))
	  (find-free-port-and-open acfg))
      (rpc:default-server-port port)
      (area-port-set! acfg port)
      (tcp-read-timeout 120000)
      ;; ((rpc:make-server (tcp-listen port)) #t)
      (tcp-listen (rpc:default-server-port)
      ))))

;; register this node by putting a packet into the pkts dir.
;; look for other servers
;; contact other servers and compile list of servers
;; there are two types of server
;;     main servers - dashboards, runners and dedicated servers - need pkt
;;     passive servers - test executers, step calls, list-runs - no pkt
;;
(define (register-node acfg hostip port-num)
  ;;(mutex-lock! (area-mutex acfg))
  (let* ((server-type  (area-server-type acfg)) ;; auto, main, passive (no pkt created)
	 (best-ip      (or hostip (get-my-best-address)))
	 (mtdir        (area-dbdir acfg))
	 (pktdir       (area-pktsdir acfg))) ;; conc mtdir "/.server-pkts")))
    (print "Registering node " best-ip ":" port-num)
    (if (not mtdir) ;; require a home for this node to put or find databases
	#f
	(begin
	  (if  (not (directory? pktdir))(create-directory pktdir))
	  ;; server is started, now create pkt if needed
	  (print "Starting server in " server-type " mode with port " port-num)
	  (if (member server-type '(auto main)) ;; TODO: if auto, count number of servers registers, if > 3 then don't put out a pkt
	      (begin
		(area-pktid-set! acfg
				 (write-alist->pkt
				  pktdir 
				  `((hostname . ,(get-host-name))
				    (ipaddr   . ,best-ip)
				    (port     . ,port-num)
				    (pid      . ,(current-process-id)))
				  pktspec: *pktspec*
				  ptype:   'server))
		(area-pktfile-set! acfg (conc pktdir "/" (area-pktid acfg) ".pkt"))))
	  (area-port-set!    acfg port-num)
	  #;(mutex-unlock! (area-mutex acfg))))))

(define *cookie-seqnum* 0)
(define (make-cookie key)
  (set! *cookie-seqnum* (add1 *cookie-seqnum*))
  ;;(print "MAKE COOKIE CALLED -- on "servkey"-"*cookie-seqnum*)
  (conc key "-" *cookie-seqnum*)
  )

;; dispatch locally if possible
;;
(define (call-deliver-response acfg ipaddr port cookie data)
  (if (and (equal? (area-myaddr acfg) ipaddr)
	   (equal? (area-port     acfg) port))
      (deliver-response acfg cookie data)
      ((rpc:procedure 'response ipaddr port) cookie data)))

(define (deliver-response acfg cookie data)
  (let ((deliver-response-start (current-milliseconds)))
    (thread-start! (make-thread
		    (lambda ()
		      (let loop ((tries-left 5))
			;;(print "TOP OF DELIVER_RESPONSE LOOP; triesleft="tries-left)
			;;(pp (hash-table->alist (area-cookie2mbox acfg)))
			(let* ((mbox (hash-table-ref/default (area-cookie2mbox acfg) cookie #f)))
			  (cond
			   ((eq? 0 tries-left)
			    (print "ulex:deliver-response: I give up. Mailbox never appeared. cookie="cookie)
			    )
			   (mbox
			    ;;(print "got mbox="mbox"  got data="data"  send.")
			    (mailbox-send! mbox data))
			   (else
			    ;;(print "no mbox yet.  look for "cookie)
			    (thread-sleep! (/ (- 6 tries-left) 10))
			    (loop (sub1 tries-left))))))
		      ;; (debug-pp (list (conc "ulex:deliver-response took " (- (current-milliseconds) deliver-response-start) " ms, cookie=" cookie " data=") data))
		      (sdbg> "deliver-response" "mailbox-send" deliver-response-start #f #f cookie)
		      )
		    (conc "deliver-response thread for cookie="cookie))))
  #t)

;; action:
;;   immediate - quick actions, no need to put in queues
;;   dbwrite   - put in dbwrite queue
;;   dbread    - put in dbread queue
;;   oslong    - os actions, e.g. du, that could take a long time
;;   osshort   - os actions that should be quick, e.g. df
;;
(define (request acfg from-ipaddr from-port servkey action cookie fname params) ;; std-peer-handler
  ;; NOTE: Use rpc:current-peer for getting return address
  (let* ((std-peer-handler-start (current-milliseconds))
	 ;; (raw-data               (alist-ref 'data     dat))
	 (rdat                   (hash-table-ref/default
				  (area-rtable acfg) action #f)) ;; this looks up the sql query or other details indexed by the action
	 (witem                  (make-witem ripaddr: from-ipaddr ;; rhost:   from-host   
					     rport:   from-port   action:  action
					     rdat:    rdat        cookie:  cookie
					     servkey: servkey     data:    params ;; TODO - rename data to params
					     caller:  (rpc:current-peer))))
    (if (not (equal? servkey (area-pktid acfg)))
	`(#f . ,(conc "I don't know you servkey=" servkey ", pktid=" (area-pktid acfg))) ;; immediately return this
	(let* ((ctype (if rdat 
			  (calldat-ctype rdat) ;; is this necessary? these should be identical
			  action)))
	  (sdbg> "std-peer-handler" "immediate" std-peer-handler-start #f #f)
	  (case ctype
	    ;; (dbwrite acfg rdat (cons from-ipaddr from-port) data)))
	    ((full-ping)  `(#t  "ack to full ping"        ,(work-queue-add acfg fname witem) ,cookie))
	    ((response)   `(#t  "ack from requestor"      ,(deliver-response acfg fname params)))
	    ((dbwrite)    `(#t  "db write submitted"      ,(work-queue-add acfg fname witem) ,cookie))
	    ((dbread)     `(#t  "db read submitted"       ,(work-queue-add acfg fname witem) ,cookie  ))
	    ((dbrw)       `(#t  "db read/write submitted" ,cookie))
	    ((osshort)    `(#t  "os short submitted"      ,cookie))
	    ((oslong)     `(#t  "os long submitted"       ,cookie))
	    (else         `(#f  "unrecognised action"     ,ctype)))))))

;; Call this to start the actual server
;;
;; start_server
;;
;;   mode: '
;;   handler: proc which takes pktrecieved as argument
;;

(define (start-server acfg)
  (let* ((conn (find-free-port-and-open acfg))
	 (port (area-port acfg)))
    (rpc:publish-procedure!
     'delist-db
     (lambda (fname)
       (hash-table-delete! (area-dbs acfg) fname)))
    (rpc:publish-procedure!
     'calling-addr
     (lambda ()
       (rpc:current-peer)))
    (rpc:publish-procedure!
     'ping
     (lambda ()(real-ping acfg)))
    (rpc:publish-procedure!
     'request
     (lambda (from-addr from-port servkey action cookie dbname params)
       (request acfg from-addr from-port servkey action cookie dbname params)))
    (rpc:publish-procedure!
     'response
     (lambda (cookie res-dat)
       (deliver-response acfg cookie res-dat)))
    (area-ready-set! acfg #t)
    (area-conn-set! acfg conn)
    ((rpc:make-server conn) #f)));; ((tcp-listen (rpc:default-server-port)) #t)


(define (launch acfg) ;;  #!optional (proc std-peer-handler))
  (print "starting launch")
  (update-known-servers acfg) ;; gotta do this on every start (thus why limit number of publicised servers)
  #;(let ((original-handler (current-exception-handler))) ;; is th
    (lambda (exception)
      (server-exit-procedure)
      (original-handler exception)))
  (on-exit (lambda ()
	     (shutdown acfg))) ;; (finalize-all-db-handles acfg)))
  ;; set up the rpc handler
  (let* ((th1  (make-thread
		(lambda ()(start-server acfg))
		"server thread"))
	 (th2   (make-thread
		 (lambda ()
		   (print "th2 starting")
		   (let loop ()
		     (work-queue-processor acfg)
		     (print "work-queue-processor crashed!")
		     (loop)))
		 "work queue thread")))
    (thread-start! th1)
    (thread-start! th2)
    (let loop ()
      (thread-sleep! 0.025)
      (if (area-ready acfg)
	  #t
	  (loop)))
    ;; attempt to fix my address
    (let* ((all-addr (get-all-ips-sorted)))	     ;; could use (tcp-addresses conn)?
      (let loop ((rem-addrs all-addr))
	(if (null? rem-addrs)
	    (begin
	      (print "ERROR: Failed to figure out the ip address of myself as a server. Giving up.")
	      (exit 1)) ;; BUG Changeme to raising an exception
		
	    (let* ((addr      (car rem-addrs))
		   (good-addr (handle-exceptions
				  exn
				  #f
				((rpc:procedure 'calling-addr addr (area-port acfg))))))
	      (if good-addr
		  (begin
		    (print "Got good-addr of " good-addr)
		    (area-myaddr-set! acfg good-addr))
		  (loop (cdr rem-addrs)))))))
    (register-node acfg (area-myaddr acfg)(area-port acfg))
    (print "INFO: Server started on " (area-myaddr acfg) ":" (area-port acfg))
    ;; (update-known-servers acfg) ;; gotta do this on every start (thus why limit number of publicised servers)
    ))

(define (clear-server-pkt acfg)
  (let ((pktf (area-pktfile acfg)))
    (if pktf (delete-file* pktf))))

(define (shutdown acfg)
  (let (;;(conn (area-conn    acfg))
	(pktf (area-pktfile acfg))
	(port (area-port    acfg)))
    (if pktf (delete-file* pktf))
    (send-all "imshuttingdown")
    ;; (rpc:close-all-connections!) ;; don't know if this is actually needed
    (finalize-all-db-handles acfg)))

(define (send-all msg)
  #f)

;; given a area record look up all the packets
;;
(define (get-all-server-pkts acfg)
  (let ((all-pkt-files (glob (conc (area-pktsdir acfg) "/*.pkt"))))
    (map (lambda (pkt-file)
	   (read-pkt->alist pkt-file pktspec: *pktspec*))
	 all-pkt-files)))

#;((Z . "9a0212302295a19610d5796fce0370fa130758e9")
  (port . "34827")
  (pid . "28748")
  (hostname . "zeus")
  (T . "server")
  (D . "1549427032.0"))

#;(define (get-my-best-address)
  (let ((all-my-addresses (get-all-ips))) ;; (vector->list (hostinfo-addresses (hostname->hostinfo (get-host-name))))))
    (cond
     ((null? all-my-addresses)
      (get-host-name))                                          ;; no interfaces?
     ((eq? (length all-my-addresses) 1)
      (ip->string (car all-my-addresses)))                      ;; only one to choose from, just go with it
     (else 
      (ip->string (car (filter (lambda (x)                      ;; take any but 127.
				 (not (eq? (u8vector-ref x 0) 127)))
			       all-my-addresses)))))))

;; whoami? I am my pkt
;;
(define (whoami? acfg)
  (hash-table-ref/default (area-hosts acfg)(area-pktid acfg) #f))

;;======================================================================
;; "Client side" operations
;;======================================================================

(define (safe-call call-key host port . params)
  (handle-exceptions
   exn
   (begin
     (print "Call " call-key " to " host ":" port " failed")
     #f)
   (apply (rpc:procedure call-key host port) params)))

;; ;; convert to/from string / sexpr
;; 
;; (define (string->sexpr str)
;;   (if (string? str)
;;       (with-input-from-string str read)
;;       str))
;; 
;; (define (sexpr->string s)
;;   (with-output-to-string (lambda ()(write s))))

;; is the server alive?
;;
(define (ping acfg host port)
  (let* ((myaddr     (area-myaddr acfg))
	 (myport     (area-port   acfg))
	 (start-time (current-milliseconds))
	 (res        (if (and (equal? myaddr host)
			      (equal? myport port))
			 (real-ping acfg)
			 ((rpc:procedure 'ping host port)))))
    (cons (- (current-milliseconds) start-time)
	  res)))

;; returns ( ipaddr port alist-fname=>randnum )
(define (real-ping acfg)
  `(,(area-myaddr acfg) ,(area-port acfg) ,(get-host-stats acfg)))

;; is the server alive AND the queues processing?
;;
#;(define (full-ping acfg servpkt)
  (let* ((start-time (current-milliseconds))
	 (res        (send-message acfg servpkt '(full-ping) 'full-ping)))
    (cons (- (current-milliseconds) start-time)
	  res))) ;; (equal? res "got ping"))))


;; look up all pkts and get the server id (the hash), port, host/ip
;; store this info in acfg
;; return the number of responsive servers found
;;
;; DO NOT VERIFY THAT THE SERVER IS ALIVE HERE. This is called at times where the current server is not yet alive and cannot ping itself
;;
(define (update-known-servers acfg)
  ;; readll all pkts
  ;; foreach pkt; if it isn't me ping the server; if alive, add to hosts hash, else rm the pkt
  (let* ((start-time (current-milliseconds))
	 (all-pkts  (delete-duplicates
		     (append (get-all-server-pkts acfg)
			     (hash-table-values (area-hosts acfg)))))
	 (hostshash (area-hosts acfg))
	 (my-id     (area-pktid acfg))
	 (pktsdir   (area-pktsdir acfg)) ;; needed to remove pkts from non-responsive servers
	 (numsrvs   0)
	 (delpkt    (lambda (pktsdir sid)
		      (print "clearing out server " sid)
		      (delete-file* (conc pktsdir "/" sid ".pkt"))
		      (hash-table-delete! hostshash sid))))
    (area-last-srvup-set! acfg (current-seconds))
    (for-each
     (lambda (servpkt)
       (if (list? servpkt)
	   ;; (pp servpkt)
	   (let* ((shost (alist-ref 'ipaddr servpkt))
		  (sport (any->number (alist-ref 'port servpkt)))
		  (res   (handle-exceptions
			  exn
			  (begin
			    ;; (print "INFO: bad server on " shost ":" sport)
			    #f)
			  (ping acfg shost sport)))
		  (sid   (alist-ref 'Z servpkt)) ;; Z code is our name for the server
		  (url   (conc shost ":" sport))
		  )
	     #;(if (or (not res)
		     (null? res))
		 (begin
		   (print "STRANGE: ping of " url " gave " res)))
	     
	     ;; (print "Got " res " from " shost ":" sport)
	     (match res
		    ((qduration . payload)
		     ;; (print "Server pkt:" (alist-ref 'ipaddr servpkt) ":" (alist-ref 'port servpkt)
		     ;;        (if payload
		     ;;            "Success" "Fail"))
		     (match payload
			    ((host port stats)
			     ;; (print "From " host ":" port " got stats: " stats)
			     (if (and host port stats)
				 (let ((url (conc host ":" port)))
				   (hash-table-set! hostshash sid servpkt)
				   ;; store based on host:port
				   (hash-table-set! (area-hoststats acfg) sid stats))
				 (print "missing data from the server, not sure what that means!"))
			     (set! numsrvs (+ numsrvs 1)))
			    (#f
			     (print "Removing pkt " sid " due to #f from server or failed ping")
			     (delpkt pktsdir sid))
			    (else
			     (print "Got ")(pp res)(print " from server ")(pp servpkt) " but response did not match (#f/#t . msg)")))
		    (else
		     ;; here we delete the pkt - can't reach the server, remove it
		     ;; however this logic is inadequate. we should mark the server as checked
		     ;; and not good, if it happens a second time - then remove the pkt
		     ;; or something similar. I.e. don't be too quick to assume the server is wedged or dead
		     ;; could be it is simply too busy to reply
		     (let ((bad-pings (hash-table-ref/default (area-health acfg) url 0)))
		       (if (> bad-pings 1) ;; two bad pings - remove pkt
			   (begin
			     (print "INFO: " bad-pings " bad responses from " url ", deleting pkt " sid)
			     (delpkt pktsdir sid))
			   (begin
			     (print "INFO: " bad-pings " bad responses from " shost ":" sport " not deleting pkt yet")
			     (hash-table-set! (area-health acfg)
					      url
					      (+ (hash-table-ref/default (area-health acfg) url 0) 1))
			     ))
		       ))))
	   ;; servpkt is not actually a pkt?
	   (begin
	     (print "Bad pkt " servpkt))))
     all-pkts)
    (sdbg> "update-known-servers" "end" start-time #f #f " found " numsrvs
	   " servers, pkts: " (map (lambda (p)
				     (alist-ref 'Z p))
				   all-pkts))
    numsrvs))

(defstruct srvstat
  (numfiles 0)   ;; number of db files handled by this server - subtract 1 for the db being currently looked at
  (randnum  #f)  ;; tie breaker number assigned to by the server itself - applies only to the db under consideration
  (pkt      #f)) ;; the server pkt

;;(define (srv->srvstat srvpkt)
  
;; Get the server best for given dbname and key
;;
;;   NOTE: key is not currently used. The key points to the kind of query, this may be useful for directing read-only queries.
;;
(define (get-best-server acfg dbname key)
  (let* (;; (servers (hash-table-values (area-hosts acfg)))
	 (servers     (area-hosts acfg))
	 (skeys       (sort (hash-table-keys servers) string>=?)) ;; a stable listing
	 (start-time  (current-milliseconds))
	 (srvstats    (make-hash-table))  ;; srvid => srvstat
	 (url         (conc (area-myaddr acfg) ":" (area-port acfg))))
    ;; (print "scores for " dbname ": " (map (lambda (k)(cons k (calc-server-score acfg dbname k))) skeys))
    (if (null? skeys)
	(if (> (update-known-servers acfg) 0)
	    (get-best-server acfg dbname key) ;; some risk of infinite loop here, TODO add try counter
	    (begin
	      (print "ERROR: no server found!") ;; since this process is also a server this should never happen
	      #f))
	(begin
	  ;; (print "in get-best-server with skeys=" skeys)
	  (if (> (- (current-seconds) (area-last-srvup acfg)) 10)
	      (begin
		(update-known-servers acfg)
		(sdbg> "get-best-server" "update-known-servers" start-time #f #f)))

	  ;; for each server look at the list of dbfiles, total number of dbs being handled
	  ;; and the rand number, save the best host
	  ;; also do a delist-db for each server dbfile not used
	  (let* ((best-server       #f)
		 (servers-to-delist (make-hash-table)))
	    (for-each
	     (lambda (srvid)
	       (let* ((server    (hash-table-ref/default servers srvid #f))
		      (stats     (hash-table-ref/default (area-hoststats acfg) srvid '(()))))
		 ;; (print "stats: " stats)
 		 (if server
		     (let* ((dbweights (car stats))
			    (srvload   (length (filter (lambda (x)(not (equal? dbname (car x)))) dbweights)))
			    (dbrec     (alist-ref dbname dbweights equal?))  ;; get the pair with fname . randscore
			    (randnum   (if dbrec
					   dbrec ;; (cdr dbrec)
					   0)))
		       (hash-table-set! srvstats srvid (make-srvstat numfiles: srvload randnum: randnum pkt: server))))))
	     skeys)
	    
	    (let* ((sorted    (sort (hash-table-values srvstats) 
				    (lambda (a b)
				      (let ((numfiles-a (srvstat-numfiles a))
					    (numfiles-b (srvstat-numfiles b))
					    (randnum-a  (srvstat-randnum a))
					    (randnum-b  (srvstat-randnum b)))
					(if (< numfiles-a numfiles-b) ;; Note, I don't think adding an offset works here. Goal was only move file handling to a different server if it has 2 less
					    #t
					    (if (and (equal? numfiles-a numfiles-b)
						     (< randnum-a randnum-b))
						#t
						#f))))))
		   (best      (if (null? sorted)
				  (begin
				    (print "ERROR: should never be null due to self as server.")
				    #f)
				  (srvstat-pkt (car sorted)))))
	      #;(print "SERVER(" url "): " dbname ": " (map (lambda (srv)
							    (let ((p (srvstat-pkt srv)))
							      (conc (alist-ref 'ipaddr p) ":" (alist-ref 'port p)
								    "(" (srvstat-numfiles srv)","(srvstat-randnum srv)")")))
							    sorted))
	      best))))))
    
    ;; send out an "I'm about to exit notice to all known servers"
    ;;
(define (death-imminent acfg)
  '())

;;======================================================================
;; U L E X  -  T H E   I N T E R E S T I N G   S T U F F ! !
;;======================================================================

;; register a handler
;;   NOTES:
;;     dbinitsql   is reserved for a list of sql statements for initializing the db
;;     dbinitfn    is reserved for a db init function, if exists called after dbinitsql
;;     
(define (register acfg key obj #!optional (ctype 'dbwrite))
  (let ((ht (area-rtable acfg)))
    (if (hash-table-exists? ht key)
	(print "WARNING: redefinition of entry " key))
    (hash-table-set! ht key (make-calldat obj: obj ctype: ctype))))

;; usage: register-batch acfg '((key1 . sql1) (key2 . sql2) ... )
;; NB// obj is often an sql query
;;
(define (register-batch acfg ctype data)
  (let ((ht (area-rtable acfg)))
    (map (lambda (dat)
	   (hash-table-set! ht (car dat)(make-calldat obj: (cdr dat) ctype: ctype)))
	 data)))

(define (initialize-area-calls-from-specfile area specfile)
  (let* ((callspec (with-input-from-file specfile read )))
    (for-each (lambda (group)
                (register-batch
                 area
                 (car group)
                 (cdr group)))
              callspec)))

;; get-rentry
;;
(define (get-rentry acfg key)
  (hash-table-ref/default (area-rtable acfg) key #f))

(define (get-rsql acfg key)
  (let ((cdat (get-rentry acfg key)))
    (if cdat
	(calldat-obj cdat)
	#f)))



;; blocking call:
;;    client                         server
;;    ------                         ------
;;    call()
;;    send-message()
;;    nmsg-send()
;;                                   nmsg-receive()
;;                                   nmsg-respond(ack,cookie)
;;    ack, cookie
;;    mbox-thread-wait(cookie)
;;                                   nmsg-send(client,cookie,result)
;;        nmsg-respond(ack)
;;        return result
;;
;; reserved action:
;;    'immediate
;;    'dbinitsql
;;
(define (call acfg dbname action params #!optional (count 0))
  (let* ((call-start-time     (current-milliseconds))
	 (srv                 (get-best-server acfg dbname action))
	 (post-get-start-time (current-milliseconds))
	 (rdat                (hash-table-ref/default (area-rtable acfg) action #f))
	 (myid                (trim-pktid (area-pktid acfg)))
	 (srvid               (trim-pktid (alist-ref 'Z srv)))
	 (cookie              (make-cookie myid)))
    (sdbg> "call" "get-best-server" call-start-time #f call-start-time " from: " myid " to server: " srvid " for " dbname " action: " action " params: " params " rdat: " rdat)
    (print "INFO: call to " (alist-ref 'ipaddr srv) ":" (alist-ref 'port srv) " from " (area-myaddr acfg) ":" (area-port acfg) " for " dbname)
    (if (and srv rdat) ;; need both to dispatch a request
	(let* ((ripaddr  (alist-ref 'ipaddr srv))
	       (rsrvid   (alist-ref 'Z srv))
	       (rport    (any->number (alist-ref 'port   srv)))
	       (res-full (if (and (equal? ripaddr (area-myaddr acfg))
				  (equal? rport   (area-port acfg)))
			     (request acfg ripaddr rport (area-pktid acfg) action cookie dbname params)
			     (safe-call 'request ripaddr rport
					(area-myaddr acfg)
					(area-port   acfg)
					#;(area-pktid acfg)
					rsrvid
					action cookie dbname params))))
	  ;; (print "res-full: " res-full)
	  (match res-full
	    ((response-ok response-msg rem ...)
	     (let* ((send-message-time (current-milliseconds))
		    ;; (match res-full
		    ;;  ((response-ok response-msg)
		    ;; (response-ok  (car res-full))
		    ;; (response-msg (cadr res-full)
		    )
	       ;; (res (take res-full 3))) ;; ctype == action, TODO: converge on one term <<=== what was this? BUG 
	       ;; (print "ulex:call: send-message took " (- send-message-time post-get-start-time) " ms params=" params)
	       (sdbg> "call" "send-message" post-get-start-time #f call-start-time)
	       (cond
		((not response-ok) #f)
		((member response-msg '("db read submitted" "db write submitted"))
		 (let* ((cookie-id   (cadddr res-full))
			(mbox        (make-mailbox))
			(mbox-time   (current-milliseconds)))
		   (hash-table-set! (area-cookie2mbox acfg) cookie-id mbox)
		   (let* ((mbox-timeout-secs    20)
			  (mbox-timeout-result 'MBOX_TIMEOUT)
			  (res                  (mailbox-receive! mbox mbox-timeout-secs mbox-timeout-result))
			  (mbox-receive-time    (current-milliseconds)))
		     (hash-table-delete! (area-cookie2mbox acfg) cookie-id)
		     (sdbg> "call" "mailbox-receive" mbox-time #f call-start-time " from: " myid " to server: " srvid " for " dbname)
		     ;; (print "ulex:call mailbox-receive took " (- mbox-receive-time mbox-time) "ms params=" params)
		     res)))
		(else
		 (print "Unhandled response \""response-msg"\"")
		 #f))
	       ;; depending on what action (i.e. ctype) is we will block here waiting for
	       ;; all the data (mechanism to be determined)
	       ;;
	       ;; if res is a "working on it" then wait
	       ;;    wait for result
	       ;; mailbox thread wait on 
	       
	       ;; if res is a "can't help you" then try a different server
	       ;; if res is a "ack" (e.g. for one-shot requests) then return res
	       ))
	    (else
	     (if (< count 10)
		 (let* ((url (conc (alist-ref 'ipaddr srv) ":" (alist-ref 'port srv))))
		   (thread-sleep! 1)
		   (print "ERROR: Bad result from " url ", dbname: " dbname ", action: " action ", params: " params ". Trying again in 1 second.")
		   (call acfg dbname action params (+ count 1)))
		 (begin
		   (error (conc "ERROR: " count " tries, still have improper response res-full=" res-full)))))))
	(begin
	  (if (not rdat)
	      (print "ERROR: action " action " not registered.")
	      (if (< count 10)
		 (begin
		   (thread-sleep! 1)
		   (area-hosts-set! acfg (make-hash-table)) ;; clear out all known hosts
		   (print "ERROR: no server found, srv=" srv ", trying again in 1 seconds")
		   (call acfg dbname action params (+ count 1)))
		 (begin
		   (error (conc "ERROR: no server found after 10 tries, srv=" srv ", giving up."))
		   #;(error "No server available"))))))))


;;======================================================================
;; U T I L I T I E S 
;;======================================================================

;; get a signature for identifing this process
;;
(define (get-process-signature)
  (cons (get-host-name)(current-process-id)))

;;======================================================================
;; S Y S T E M   S T U F F
;;======================================================================

;; get normalized cpu load by reading from /proc/loadavg and
;; /proc/cpuinfo return all three values and the number of real cpus
;; and the number of threads returns alist '((adj-cpu-load
;; . normalized-proc-load) ... etc.  keys: adj-proc-load,
;; adj-core-load, 1m-load, 5m-load, 15m-load
;;
(define (get-normalized-cpu-load)
  (let ((res (get-normalized-cpu-load-raw))
	(default `((adj-proc-load . 2) ;; there is no right answer
		   (adj-core-load . 2)
		   (1m-load       . 2)
		   (5m-load       . 0) ;; causes a large delta - thus causing default of throttling if stuff goes wrong
		   (15m-load      . 0)
		   (proc          . 1)
		   (core          . 1)
		   (phys          . 1)
		   (error         . #t))))
    (cond
     ((and (list? res)
	   (> (length res) 2))
      res)
     ((eq? res #f)   default) ;; add messages?
     ((eq? res #f) default)   ;; this would be the #eof
     (else default))))

(define (get-normalized-cpu-load-raw)
  (let* ((actual-host           (get-host-name))) ;; #f is localhost
    (let ((data  (append 
		  (with-input-from-file "/proc/loadavg" read-lines)
		  (with-input-from-file "/proc/cpuinfo" read-lines)
		  (list "end")))
	  (load-rx  (regexp "^([\\d\\.]+)\\s+([\\d\\.]+)\\s+([\\d\\.]+)\\s+.*$"))
	  (proc-rx  (regexp "^processor\\s+:\\s+(\\d+)\\s*$"))
	  (core-rx  (regexp "^core id\\s+:\\s+(\\d+)\\s*$"))
	  (phys-rx  (regexp "^physical id\\s+:\\s+(\\d+)\\s*$"))
	  (max-num  (lambda (p n)(max (string->number p) n))))
      ;; (print "data=" data)
      (if (null? data) ;; something went wrong
	  #f
	  (let loop ((hed      (car data))
		     (tal      (cdr data))
		     (loads    #f)
		     (proc-num 0)  ;; processor includes threads
		     (phys-num 0)  ;; physical chip on motherboard
		     (core-num 0)) ;; core
	    ;; (print hed ", " loads ", " proc-num ", " phys-num ", " core-num)
	    (if (null? tal) ;; have all our data, calculate normalized load and return result
		(let* ((act-proc (+ proc-num 1))
		       (act-phys (+ phys-num 1))
		       (act-core (+ core-num 1))
		       (adj-proc-load (/ (car loads) act-proc))
		       (adj-core-load (/ (car loads) act-core))
		       (result
			(append (list (cons 'adj-proc-load adj-proc-load)
				      (cons 'adj-core-load adj-core-load))
				(list (cons '1m-load (car loads))
				      (cons '5m-load (cadr loads))
				      (cons '15m-load (caddr loads)))
				(list (cons 'proc act-proc)
				      (cons 'core act-core)
				      (cons 'phys act-phys)))))
		  result)
		(regex-case
		    hed
		  (load-rx  ( x l1 l5 l15 ) (loop (car tal)(cdr tal)(map string->number (list l1 l5 l15)) proc-num phys-num core-num))
		  (proc-rx  ( x p         ) (loop (car tal)(cdr tal) loads           (max-num p proc-num) phys-num core-num))
		  (phys-rx  ( x p         ) (loop (car tal)(cdr tal) loads           proc-num (max-num p phys-num) core-num))
		  (core-rx  ( x c         ) (loop (car tal)(cdr tal) loads           proc-num phys-num (max-num c core-num)))
		  (else 
		   (begin
		     ;; (print "NO MATCH: " hed)
		     (loop (car tal)(cdr tal) loads proc-num phys-num core-num))))))))))

(define (get-host-stats acfg)
  (let ((stats-hash (area-stats acfg)))
    ;; use this opportunity to remove references to dbfiles which have not been accessed in a while
    (for-each
     (lambda (dbname)
       (let* ((stats       (hash-table-ref stats-hash dbname))
	      (last-access (stat-when stats)))
	 (if (and (> last-access 0)                             ;; if zero then there has been no access
		  (> (- (current-seconds) last-access) 10))     ;; not used in ten seconds
	     (begin
	       (print "Removing " dbname " from stats list")
	       (hash-table-delete! stats-hash dbname) ;; remove from stats hash
	       (stat-dbs-set! stats (hash-table-keys stats))))))
     (hash-table-keys stats-hash))
    
    `(,(hash-table->alist (area-dbs acfg)) ;; dbname => randnum
      ,(map (lambda (dbname)  ;; dbname is the db name
	      (cons dbname (stat-when (hash-table-ref stats-hash dbname))))
	    (hash-table-keys stats-hash))
      (cpuload . ,(get-normalized-cpu-load)))))
    #;(stats   . ,(map (lambda (k) ;; create an alist from the stats data
		       (cons k (stat->alist (hash-table-ref (area-stats acfg) k))))
		     (hash-table-keys (area-stats acfg))))

#;(trace
 ;; assv
 ;; cdr
 ;; caar
 ;; ;; cdr
 ;; call
 ;; finalize-all-db-handles
 ;; get-all-server-pkts
 ;; get-normalized-cpu-load
 ;; get-normalized-cpu-load-raw
 ;; launch
 ;; nmsg-send
 ;; process-db-queries
 ;; receive-message
 ;; std-peer-handler
 ;; update-known-servers
 ;; work-queue-processor
 )

;;======================================================================
;; netutil
;;   move this back to ulex-netutil.scm someday?
;;======================================================================

;; #include <stdio.h>
;; #include <netinet/in.h>
;; #include <string.h>
;; #include <arpa/inet.h>

(foreign-declare "#include \"sys/types.h\"")
(foreign-declare "#include \"sys/socket.h\"")
(foreign-declare "#include \"ifaddrs.h\"")
(foreign-declare "#include \"arpa/inet.h\"")

;; get IP addresses from ALL interfaces
(define get-all-ips
  (foreign-safe-lambda* scheme-object ()
    "

// from https://stackoverflow.com/questions/17909401/linux-c-get-default-interfaces-ip-address :


    C_word lst = C_SCHEME_END_OF_LIST, len, str, *a;
//    struct ifaddrs *ifa, *i;
//    struct sockaddr *sa;

    struct ifaddrs * ifAddrStruct = NULL;
    struct ifaddrs * ifa = NULL;
    void * tmpAddrPtr = NULL;

    if ( getifaddrs(&ifAddrStruct) != 0)
      C_return(C_SCHEME_FALSE);

//    for (i = ifa; i != NULL; i = i->ifa_next) {
    for (ifa = ifAddrStruct; ifa != NULL; ifa = ifa->ifa_next) {
        if (ifa->ifa_addr->sa_family==AF_INET) { // Check it is
            // a valid IPv4 address
            tmpAddrPtr = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
            char addressBuffer[INET_ADDRSTRLEN];
            inet_ntop(AF_INET, tmpAddrPtr, addressBuffer, INET_ADDRSTRLEN);
//            printf(\"%s IP Address %s\\n\", ifa->ifa_name, addressBuffer);
            len = strlen(addressBuffer);
            a = C_alloc(C_SIZEOF_PAIR + C_SIZEOF_STRING(len));
            str = C_string(&a, len, addressBuffer);
            lst = C_a_pair(&a, str, lst);
        } 

//        else if (ifa->ifa_addr->sa_family==AF_INET6) { // Check it is
//            // a valid IPv6 address
//            tmpAddrPtr = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
//            char addressBuffer[INET6_ADDRSTRLEN];
//            inet_ntop(AF_INET6, tmpAddrPtr, addressBuffer, INET6_ADDRSTRLEN);
////            printf(\"%s IP Address %s\\n\", ifa->ifa_name, addressBuffer);
//            len = strlen(addressBuffer);
//            a = C_alloc(C_SIZEOF_PAIR + C_SIZEOF_STRING(len));
//            str = C_string(&a, len, addressBuffer);
//            lst = C_a_pair(&a, str, lst);
//       }

//       else {
//         printf(\" not an IPv4 address\\n\");
//       }

    }

    freeifaddrs(ifa);
    C_return(lst);

"))

;; Change this to bias for addresses with a reasonable broadcast value?
;;
(define (ip-pref-less? a b)
  (let* ((rate (lambda (ipstr)
                 (regex-case ipstr
                             ( "^127\\." _ 0 )
                             ( "^(10\\.0|192\\.168\\.)\\..*" _ 1 )
                             ( else 2 ) ))))
    (< (rate a) (rate b))))
  

(define (get-my-best-address)
  (let ((all-my-addresses (get-all-ips))
        ;;(all-my-addresses-old (vector->list (hostinfo-addresses (hostname->hostinfo (get-host-name)))))
        )
    (cond







|


<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<

|




>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<

<
<
<
<
<
<

|

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<

|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<




<
<
<
<
<
|







22
23
24
25
26
27
28
29
30
31




























32


33
34
35
36
37
38
39
40



































































































41























































































































































































































































































































































































































42






43
44
45




























































































































































































































































































































































































































































































































































































































































































46
47




































48









49














50
























































































51









52
53
54
55





56
57
58
59
60
61
62
63
;;   See README in the distribution at https://www.kiatoa.com/fossils/ulex
;; NOTES:
;;   Why sql-de-lite and not say, dbi?  - performance mostly, then simplicity.
;;
;;======================================================================

;; (use rpc pkts mailbox sqlite3)

(module ulex
    *































(import scheme posix chicken data-structures ports extras files mailbox)
(import srfi-18 pkts matchable regex
	typed-records srfi-69 srfi-1
	srfi-4 regex-case
	(prefix sqlite3 sqlite3:)
	foreign
	tcp6
	;; ulex-netutil



































































































	hostinfo)






























































































































































































































































































































































































































;;======================================================================
;; network utilities
;;======================================================================





























































































































































































































































































































































































































































































































































































































































































(define (rate-ip ipaddr)




































  (regex-case ipaddr









    ( "^127\\..*" _ 0 )














    ( "^(10\\.0|192\\.168)\\..*" _ 1 )
























































































    ( else 2 ) ))










;; Change this to bias for addresses with a reasonable broadcast value?
;;
(define (ip-pref-less? a b)





  (> (rate-ip a) (rate-ip b)))
  

(define (get-my-best-address)
  (let ((all-my-addresses (get-all-ips))
        ;;(all-my-addresses-old (vector->list (hostinfo-addresses (hostname->hostinfo (get-host-name)))))
        )
    (cond
1444
1445
1446
1447
1448
1449
1450


































































































































































































































































































































1451


































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































1452
     ;;    		       all-my-addresses))))

     )))

(define (get-all-ips-sorted)
  (sort (get-all-ips) ip-pref-less?))






































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































)







>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>

>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
     ;;    		       all-my-addresses))))

     )))

(define (get-all-ips-sorted)
  (sort (get-all-ips) ip-pref-less?))

(define (get-all-ips)
  (map ip->string (vector->list 
		   (hostinfo-addresses
		    (host-information (current-hostname))))))

;; make it a global? Well, it is local to area module

(define *captain-pktspec*
  `((captain (host     . h)
	     (port     . p)
	     (pid      . i)
	     (ipaddr   . a)
	     )
    #;(data   (hostname . h)  ;; sender hostname
	    (port     . p)  ;; sender port
	    (ipaddr   . a)  ;; sender ip
	    (hostkey  . k)  ;; sending host key - store info at server under this key
	    (servkey  . s)  ;; server key - this needs to match at server end or reject the msg
	    (format   . f)  ;; sb=serialized-base64, t=text, sx=sexpr, j=json
	    (data     . d)  ;; base64 encoded slln data
	    )))

;; struct for keeping track of our world

(defstruct udat
  (captain-address #f)
  (captain-host    #f)
  (captain-port    #f)
  (captain-pid     #f)
  (cpkts-dir       (conc (get-environment-variable "HOME") "/.ulex/pkts"))
  (cpkt-spec       *captain-pktspec*)
  (my-cpkt-key     #f)   ;; put Z card here when I create a pkt for myself as captain
  (my-address      #f)
  (my-hostname     #f)
  (my-port         #f)
  (my-pid          (current-process-id))
  (serv-listener   #f)
  (handler-thread  #f)
  (handlers        (make-hash-table))
  (outgoing-conns  (make-hash-table))  ;; host:port -> conn
  )

;; struct for keeping track of others we are talking to

(defstruct peer
  (addr-port       #f)
  (hostname        #f)
  (pid             #f)
  (inp             #f)  ;; input port from the peer
  (oup             #f)  ;; output port to the peer
  (owns            '()) ;; list of databases this peer is currently handling
  )

;;======================================================================
;; Captain pkt functions
;;======================================================================

;; given a pkts dir read 
;;
(define (get-all-captain-pkts udata)
  (let* ((pktsdir       (let ((d (udat-cpkts-dir udata)))
			  (if (file-exists? d)
			      d
			      (begin
				(create-directory d #t)
				d))))
	 (all-pkt-files (glob (conc pktsdir "/*.pkt")))
	 (pkt-spec      (udat-cpkt-spec udata)))
    (map (lambda (pkt-file)
	   (read-pkt->alist pkt-file pktspec: pkt-spec))
	 all-pkt-files)))

;; sort by D then Z, return one, choose the oldest then
;; differentiate if needed using the Z key
;;l
(define (get-winning-pkt pkts)
  (if (null? pkts)
      #f
      (car (sort pkts (lambda (a b)
			(let ((ad (string->number (alist-ref 'D a)))
			      (bd (string->number (alist-ref 'D b))))
			  (if (eq? a b)
			      (let ((az (alist-ref 'Z a))
				    (bz (alist-ref 'Z b)))
				(string>=? az bz))
			      (> ad bd))))))))

;; create a tcp listener and return a populated udat struct with
;; my port, address, hostname, pid etc.
;; return #f if fail to find a port to allocate.
;;
(define (start-server-find-port udata #!optional (port 4242)) 
  (handle-exceptions
      exn
      (if (< port 65535)(start-server-find-port udata (+ port 1)) #f)
    (connect-server udata port)))

(define (connect-server udata port)
  ;; (tcp-listener-socket LISTENER)(socket-name so)
  ;; sockaddr-address, sockaddr-port, sockaddr->string
  (let* ((tlsn (tcp-listen port 1000 #f)) ;; (tcp-listen TCPPORT [BACKLOG [HOST]])
	 (addr (get-my-best-address))) ;; (hostinfo-addresses (host-information (current-hostname)))
    (udat-my-address-set!    udata addr)
    (udat-my-port-set!       udata port)
    (udat-my-hostname-set!   udata (get-host-name))
    (udat-serv-listener-set! udata tlsn)
    udata))

;; put the host, ip, port and pid into a pkt in
;; the captain pkts dir
;;  - assumes user has already fired up a server
;;    which will be in the udata struct
;;
(define (create-captain-pkt udata)
  (if (not (udat-serv-listener udata))
      (begin
	(print "ERROR: create-captain-pkt called with out a listener")
	#f)
      (let* ((pktdat `((port   . ,(udat-my-port udata))
		       (host   . ,(udat-my-hostname udata))
		       (ipaddr . ,(udat-my-address udata))
		       (pid    . ,(udat-my-pid     udata))))
	     (pktdir  (udat-cpkts-dir udata))
	     (pktspec (udat-cpkt-spec udata))
	     )
	(udat-my-cpkt-key-set!
	 udata
	 (write-alist->pkt
	  pktdir
	  pktdat
	  pktspec: pktspec
	  ptype:   'captain))
	(udat-my-cpkt-key udata))))

;; NB// This needs to be started in a thread
;;
;; setup to be a captain
;;   - start server
;;   - create pkt
;;   - start server port handler
;;
(define (setup-as-captain udata)
  (if (start-server-find-port udata) ;; puts the server in udata
      (if (create-captain-pkt udata)
	  (let* ((th (make-thread (lambda ()
				    (ulex-handler udata)) "Captain handler")))
	    (udat-handler-thread-set! udata th)
	    (thread-start! th))
	  #f)
      #f))

(define (get-peer-dat udata host-port #!optional (hostname #f)(pid #f))
  (let* ((pdat (or (hash-table-ref/default (udat-outgoing-conns udata) host-port #f)
		   (let ((npdat (make-peer addr-port: host-port)))
		     (if hostname (peer-hostname-set! npdat hostname))
		     (if pid (peer-pid-set! npdat pid))
		     (let-values (((ninp noup)(tcp-connect host-port)))
		       (peer-inp-set! npdat ninp)
		       (peer-oup-set! npdat noup))
		     (hash-table-set! (udat-outgoing-conns udata) host-port npdat)
		     npdat))))
    pdat))

(define (get-peer-ports udata host-port #!optional (hostname #f)(pid #f))
  (let ((pdat (get-peer-dat udata host-port hostname pid)))
    (values (peer-inp pdat)(peer-oup pdat))))

;; send back ack, amusing I suppose that this looks almost like what
;; tcp itself does ...
;;
(define (reply udata host-port handler qrykey data #!optional (hostname #f)(pid #f))
  (let-values (((inp oup)(get-peer-ports udata host-port hostname pid)))
    (write-line (conc
		 handler " "
		 (udat-my-address  udata) ":" (udat-my-port udata) " "
		 (udat-my-hostname udata) " "
		 (udat-my-pid      udata) " "
		 qrykey)
		oup)
    (write-line data oup))) ;; we must send a second line - for the ack let it be the qrykey 

(define (add-to-work-queue udata . blah)
  #f)

;; send back ack
;;
(define (send-ack udata host-port qrykey #!optional (hostname #f)(pid #f))
  (reply udata "ack" qrykey qrykey hostname pid)) ;; we must send a second line - for the ack let it be the qrykey 
  
;; 
;;
(define (ulex-handler udata)
  (let* ((serv-listener (udat-serv-listener udata)))
    (let-values (((inp oup)(tcp-accept serv-listener)))
      ;; data comes as two lines
      ;;   handlerkey resp-addr:resp-port hostname pid qrykey [dbpath/dbfile.db]
      ;;   data
      (let loop ((state 'start))
	(let* ((controldat (read-line inp))
	       (data       (read-line inp)))
	  (match (string-split controldat)
	    ((handlerkey host:port hostname pid qrykey params ...)
	     (case (string->symbol handlerkey)
	       ((ack)(print "Got ack!"))
	       ((ping)
		(let* ((proc (hash-table-ref/default (udat-handlers udata) 'ping #f))
		       (val  (if proc (proc) "gotping")))
		  (reply udata host:port "version" qrykey val)))
	       ((rucaptain)
		(reply udata host:port "iamcaptain" qrykey (if (udat-my-cpkt-key udata)
							       "yes"
							       "no")))
	       (else
		(send-ack udata host:port qrykey hostname pid)
		(add-to-work-queue udata (get-peer-dat udata host:port) handlerkey data)))
	     (else (print "BAD DATA? handler=" handlerkey " data=" data)))))
	(loop state)))))

;; add a proc to the handler list
(define (register-handler udata key proc)
  (hash-table-set! (udat-handlers udata) key proc))

;;======================================================================
;; connection setup and management functions
;;======================================================================

;; find or become the captain, return a ulex object
;;
(define (setup)
  (let* ((udata (make-udat))
	 (cpkts (get-all-captain-pkts udata)) ;; read captain pkts
	 (captn (get-winning-pkt cpkts)))
    (if captn
	(let* ((port   (alist-ref 'port   captn))
	       (host   (alist-ref 'host   captn))
	       (ipaddr (alist-ref 'ipaddr captn))
	       (pid    (alist-ref 'pid    captn)))
	  (udat-captain-address-set! udata ipaddr)
	  (udat-captain-host-set!    udata host)
	  (udat-captain-port-set!    udata port)
	  (udat-captain-pid-set!     udata pid)
	  ;;(if (ping-captain udata)
	  ;;    udata
	  ;;    (begin
	  ;;       (remove-captain-pkt udata captn)
	  ;;       (setup)))
	  udata)
	(setup-as-captain udata)) ;; this saves the thread to captain-thread and starts the thread
    ))
    
(define (connect udata dbfname)
  udata)

) ;; END OF ULEX


;;; ;;======================================================================
;;; ;; D E B U G   H E L P E R S
;;; ;;======================================================================
;;;     
;;; (define (dbg> . args)
;;;   (with-output-to-port (current-error-port)
;;;     (lambda ()
;;;       (apply print "dbg> " args))))
;;; 
;;; (define (debug-pp . args)
;;;   (if (get-environment-variable "ULEX_DEBUG")
;;;       (with-output-to-port (current-error-port)
;;; 	(lambda ()
;;; 	  (apply pp args)))))
;;; 
;;; (define *default-debug-port* (current-error-port))
;;; 
;;; (define (sdbg> fn stage-name stage-start stage-end start-time . message)
;;;   (if (get-environment-variable "ULEX_DEBUG")
;;;       (with-output-to-port *default-debug-port* 
;;; 	(lambda ()
;;; 	  (apply print "ulex:" fn " " stage-name " took " (- (if stage-end stage-end (current-milliseconds)) stage-start) " ms. "
;;; 		 (if start-time
;;; 		     (conc "total time " (- (current-milliseconds) start-time)
;;; 			   " ms.")
;;; 		     "")
;;; 		 message
;;; 		 )))))

;;======================================================================
;; M A C R O S
;;======================================================================
;; iup callbacks are not dumping the stack, this is a work-around
;;

;; Some of these routines use:
;;
;;     http://www.cs.toronto.edu/~gfb/scheme/simple-macros.html
;;
;; Syntax for defining macros in a simple style similar to function definiton,
;;  when there is a single pattern for the argument list and there are no keywords.
;;
;; (define-simple-syntax (name arg ...) body ...)
;;
;; 
;; (define-syntax define-simple-syntax
;;   (syntax-rules ()
;;     ((_ (name arg ...) body ...)
;;      (define-syntax name (syntax-rules () ((name arg ...) (begin body ...)))))))
;; 
;; (define-simple-syntax (catch-and-dump proc procname)
;;   (handle-exceptions
;;    exn
;;    (begin
;;      (print-call-chain (current-error-port))
;;      (with-output-to-port (current-error-port)
;;        (lambda ()
;;          (print ((condition-property-accessor 'exn 'message) exn))
;;          (print "Callback error in " procname)
;;          (print "Full condition info:\n" (condition->list exn)))))
;;    (proc)))
;; 
;; 
;;======================================================================
;;  R E C O R D S
;;======================================================================

;;; ;; information about me as a server
;;; ;;
;;; (defstruct area
;;;   ;; about this area
;;;   (useportlogger #f)
;;;   (lowport       32768)
;;;   (server-type   'auto)  ;; auto=create up to five servers/pkts, main=create pkts, passive=no pkt (unless there are no pkts at all)
;;;   (conn          #f)
;;;   (port          #f)
;;;   (myaddr        (get-my-best-address))
;;;   pktid          ;; get pkt from hosts table if needed
;;;   pktfile
;;;   pktsdir
;;;   dbdir
;;;   (dbhandles     (make-hash-table)) ;; fname => list-of-dbh, NOTE: Should really never need more than one?
;;;   (mutex         (make-mutex))
;;;   (rtable        (make-hash-table)) ;; registration table of available actions
;;;   (dbs           (make-hash-table)) ;; filename => random number, used for choosing what dbs I serve
;;;   ;; about other servers
;;;   (hosts         (make-hash-table)) ;; key => hostdat
;;;   (hoststats     (make-hash-table)) ;; key => alist of fname => ( qcount . qtime )
;;;   (reqs          (make-hash-table)) ;; uri => queue
;;;   ;; work queues
;;;   (wqueues       (make-hash-table)) ;; fname => qdat
;;;   (stats         (make-hash-table)) ;; fname => totalqueries
;;;   (last-srvup    (current-seconds)) ;; last time we updated the known servers
;;;   (cookie2mbox   (make-hash-table)) ;; map cookie for outstanding request to mailbox of awaiting call
;;;   (ready #f)
;;;   (health        (make-hash-table)) ;; ipaddr:port => num failed pings since last good ping
;;;   )
;;; 
;;; ;; host stats
;;; ;;
;;; (defstruct hostdat
;;;   (pkt      #f)
;;;   (dbload   (make-hash-table))  ;; "dbfile.db" => queries/min
;;;   (hostload #f)                 ;; normalized load ( 5min load / numcpus )
;;;   )
;;; 
;;; ;; dbdat
;;; ;;
;;; (defstruct dbdat
;;;   (dbh    #f)
;;;   (fname  #f)
;;;   (write-access #f)
;;;   (sths   (make-hash-table))  ;; hash mapping query strings to handles
;;;   )
;;; 
;;; ;; qdat
;;; ;;
;;; (defstruct qdat
;;;   (writeq  (make-queue))
;;;   (readq   (make-queue))
;;;   (rwq     (make-queue))
;;;   (logq    (make-queue)) ;; do we need a queue for logging? yes, if we use sqlite3 db for logging
;;;   (osshort (make-queue))
;;;   (oslong  (make-queue))
;;;   (misc    (make-queue)) ;; used for things like ping-full
;;;   )
;;; 
;;; ;; calldat
;;; ;;
;;; (defstruct calldat
;;;   (ctype 'dbwrite)
;;;   (obj   #f)              ;; this would normally be an SQL statement e.g. SELECT, INSERT etc.
;;;   (rtime (current-milliseconds)))
;;; 
;;; ;; make it a global? Well, it is local to area module
;;; 
;;; (define *pktspec*
;;;   `((server (hostname . h)
;;; 	    (port     . p)
;;; 	    (pid      . i)
;;; 	    (ipaddr   . a)
;;; 	    )
;;;     (data   (hostname . h)  ;; sender hostname
;;; 	    (port     . p)  ;; sender port
;;; 	    (ipaddr   . a)  ;; sender ip
;;; 	    (hostkey  . k)  ;; sending host key - store info at server under this key
;;; 	    (servkey  . s)  ;; server key - this needs to match at server end or reject the msg
;;; 	    (format   . f)  ;; sb=serialized-base64, t=text, sx=sexpr, j=json
;;; 	    (data     . d)  ;; base64 encoded slln data
;;; 	    )))
;;; 
;;; ;; work item
;;; ;;
;;; (defstruct witem
;;;   (rhost #f)   ;; return host
;;;   (ripaddr #f) ;; return ipaddr
;;;   (rport #f)   ;; return port
;;;   (servkey #f) ;; the packet representing the client of this workitem, used by final send-message
;;;   (rdat  #f)   ;; the request - usually an sql query, type is rdat
;;;   (action #f)  ;; the action: immediate, dbwrite, dbread,oslong, osshort
;;;   (cookie #f)  ;; cookie id for response
;;;   (data   #f)  ;; the data payload, i.e. parameters
;;;   (result #f)  ;; the result from processing the data
;;;   (caller #f)) ;; the calling peer according to rpc itself
;;; 
;;; (define (trim-pktid pktid)
;;;   (if (string? pktid)
;;;       (substring pktid 0 4)
;;;       "nopkt"))
;;; 
;;; (define (any->number num)
;;;   (cond
;;;    ((number? num) num)
;;;    ((string? num) (string->number num))
;;;    (else num)))
;;; 
;;; (use trace)
;;; (trace-call-sites #t)
;;; 
;;; ;;======================================================================
;;; ;; D A T A B A S E   H A N D L I N G 
;;; ;;======================================================================
;;; 
;;; ;; look in dbhandles for a db, return it, else return #f
;;; ;;
;;; (define (get-dbh acfg fname)
;;;   (let ((dbh-lst (hash-table-ref/default (area-dbhandles acfg) fname '())))
;;;     (if (null? dbh-lst)
;;; 	(begin
;;; 	  ;; (print "opening db for " fname)
;;; 	  (open-db acfg fname)) ;; Note that the handles get put back in the queue in the save-dbh calls
;;; 	(let ((rem-lst (cdr dbh-lst)))
;;; 	  ;; (print "re-using saved connection for " fname)
;;; 	  (hash-table-set! (area-dbhandles acfg) fname rem-lst)
;;; 	  (car dbh-lst)))))
;;; 
;;; (define (save-dbh acfg fname dbdat)
;;;     ;; (print "saving dbh for " fname)
;;;     (hash-table-set! (area-dbhandles acfg) fname (cons dbdat (hash-table-ref/default (area-dbhandles acfg) fname '()))))
;;; 
;;; ;; open the database, if never before opened init it. put the handle in the
;;; ;; open db's hash table
;;; ;; returns: the dbdat
;;; ;;
;;; (define (open-db acfg fname)
;;;   (let* ((fullname     (conc (area-dbdir acfg) "/" fname))
;;; 	 (exists       (file-exists? fullname))
;;; 	 (write-access (if exists
;;; 			   (file-write-access? fullname)
;;; 			   (file-write-access? (area-dbdir acfg))))
;;; 	 (db           (sqlite3:open-database fullname))
;;; 	 (handler      (sqlite3:make-busy-timeout 136000))
;;; 	 )
;;;     (sqlite3:set-busy-handler! db handler)
;;;     (sqlite3:execute db "PRAGMA synchronous = 0;")
;;;     (if (not exists) ;; need to init the db
;;; 	(if write-access
;;; 	    (let ((isql (get-rsql acfg 'dbinitsql))) ;; get the init sql statements
;;; 	      ;; (sqlite3:with-transaction
;;; 	      ;;  db
;;; 	      ;;  (lambda ()
;;; 		 (if isql
;;; 		     (for-each
;;; 		      (lambda (sql)
;;; 			(sqlite3:execute db sql))
;;; 		      isql)))
;;; 	    (print "ERROR: no write access to " (area-dbdir acfg))))
;;;     (make-dbdat dbh: db fname: fname write-access: write-access)))
;;; 
;;; ;; This is a low-level command to retrieve or to prepare, save and return a prepared statment
;;; ;; you must extract the db handle
;;; ;;
;;; (define (get-sth db cache stmt)
;;;   (if (hash-table-exists? cache stmt)
;;;       (begin
;;; 	;; (print "Reusing cached stmt for " stmt)
;;; 	(hash-table-ref/default cache stmt #f))
;;;       (let ((sth (sqlite3:prepare db stmt)))
;;; 	(hash-table-set! cache stmt sth)
;;; 	;; (print "prepared stmt for " stmt)
;;; 	sth)))
;;; 
;;; ;; a little more expensive but does all the tedious deferencing - only use if you don't already
;;; ;; have dbdat and db sitting around
;;; ;;
;;; (define (full-get-sth acfg fname stmt)
;;;   (let* ((dbdat  (get-dbh acfg fname))
;;; 	 (db     (dbdat-dbh dbdat))
;;; 	 (sths   (dbdat-sths dbdat)))
;;;     (get-sth db sths stmt)))
;;; 
;;; ;; write to a db
;;; ;; acfg: area data
;;; ;; rdat: request data
;;; ;; hdat: (host . port)
;;; ;;
;;; ;; (define (dbwrite acfg rdat hdat data-in)
;;; ;;   (let* ((dbname (car data-in))
;;; ;; 	 (dbdat  (get-dbh acfg dbname))
;;; ;; 	 (db     (dbdat-dbh dbdat))
;;; ;; 	 (sths   (dbdat-sths dbdat))
;;; ;; 	 (stmt   (calldat-obj rdat))
;;; ;; 	 (sth    (get-sth db sths stmt))
;;; ;; 	 (data   (cdr data-in)))
;;; ;;     (print "dbname: " dbname " acfg: " acfg " rdat: " (calldat->alist rdat) " hdat: " hdat " data: " data)
;;; ;;     (print "dbdat: " (dbdat->alist dbdat))
;;; ;;     (apply sqlite3:execute sth data)
;;; ;;     (save-dbh acfg dbname dbdat)
;;; ;;     #t
;;; ;;     ))
;;; 
;;; (define (finalize-all-db-handles acfg)
;;;   (let* ((dbhandles (area-dbhandles acfg))  ;; dbhandles is hash of fname ==> dbdat
;;; 	 (num       0))
;;;     (for-each
;;;      (lambda (area-name)
;;;        (print "Closing handles for " area-name)
;;;        (let ((dbdats (hash-table-ref/default dbhandles area-name '())))
;;; 	 (for-each
;;; 	  (lambda (dbdat)
;;; 	    ;; first close all statement handles
;;; 	    (for-each
;;; 	     (lambda (sth)
;;; 	       (sqlite3:finalize! sth)
;;; 	       (set! num (+ num 1)))
;;; 	     (hash-table-values (dbdat-sths dbdat)))
;;; 	    ;; now close the dbh
;;; 	    (set! num (+ num 1))
;;; 	    (sqlite3:finalize! (dbdat-dbh dbdat)))
;;; 	  dbdats)))
;;;      (hash-table-keys dbhandles))
;;;     (print "FINALIZED " num " dbhandles")))
;;; 
;;; ;;======================================================================
;;; ;; W O R K   Q U E U E   H A N D L I N G 
;;; ;;======================================================================
;;; 
;;; (define (register-db-as-mine acfg dbname)
;;;   (let ((ht (area-dbs acfg)))
;;;     (if (not (hash-table-ref/default ht dbname #f))
;;; 	(hash-table-set! ht dbname (random 10000)))))
;;; 	
;;; (define (work-queue-add acfg fname witem)
;;;   (let* ((work-queue-start (current-milliseconds))
;;; 	 (action           (witem-action witem)) ;; NB the action is the index into the rdat actions
;;; 	 (qdat             (or (hash-table-ref/default (area-wqueues acfg) fname #f)
;;; 			       (let ((newqdat (make-qdat)))
;;; 				 (hash-table-set! (area-wqueues acfg) fname newqdat)
;;; 				 newqdat)))
;;; 	 (rdat             (hash-table-ref/default (area-rtable acfg) action #f)))
;;;     (if rdat
;;; 	(queue-add!
;;; 	 (case (calldat-ctype rdat)
;;; 	   ((dbwrite)   (register-db-as-mine acfg fname)(qdat-writeq qdat))
;;; 	   ((dbread)    (register-db-as-mine acfg fname)(qdat-readq  qdat))
;;; 	   ((dbrw)      (register-db-as-mine acfg fname)(qdat-rwq    qdat))
;;; 	   ((oslong)    (qdat-oslong qdat))
;;; 	   ((osshort)   (qdat-osshort qdat))
;;; 	   ((full-ping) (qdat-misc  qdat))
;;; 	   (else
;;; 	    (print "ERROR: no queue for " action ". Adding to dbwrite queue.")
;;; 	    (qdat-writeq qdat)))
;;; 	 witem)
;;; 	(case action
;;; 	  ((full-ping)(qdat-misc qdat))
;;; 	  (else
;;; 	   (print "ERROR: No action " action " was registered"))))
;;;     (sdbg> "work-queue-add" "queue-add" work-queue-start #f #f)
;;;     #t)) ;; for now, simply return #t to indicate request got to the queue
;;; 
;;; (define (doqueue acfg q fname dbdat dbh)
;;;   ;; (print "doqueue: " fname)
;;;   (let* ((start-time (current-milliseconds))
;;; 	 (qlen       (queue-length q)))
;;;     (if (> qlen 1)
;;; 	(print "Processing queue of length " qlen))
;;;     (let loop ((count      0)
;;; 	       (responses '()))
;;;       (let ((delta (- (current-milliseconds) start-time)))
;;; 	(if (or (queue-empty? q)
;;; 		(> delta 400)) ;; stop working on this queue after 400ms have passed
;;; 	    (list count delta responses) ;; return count, delta and responses list
;;; 	    (let* ((witem  (queue-remove! q))
;;; 		   (action (witem-action witem))
;;; 		   (rdat   (witem-rdat   witem))
;;; 		   (stmt   (calldat-obj rdat))
;;; 		   (sth    (full-get-sth acfg fname stmt))
;;; 		   (ctype  (calldat-ctype rdat))
;;; 		   (data   (witem-data   witem))
;;; 		   (cookie (witem-cookie witem)))
;;; 	      ;; do the processing and save the result in witem-result
;;; 	      (witem-result-set!
;;; 	       witem
;;; 	       (case ctype ;; action
;;; 		 ((noblockwrite) ;; blind write, no ack of success returned
;;; 		  (apply sqlite3:execute sth data)
;;; 		  (sqlite3:last-insert-rowid dbh))
;;; 		 ((dbwrite)      ;; blocking write   
;;; 		  (apply sqlite3:execute sth data)
;;; 		  #t)
;;; 		 ((dbread) ;; TODO: consider breaking this up and shipping in pieces for large query
;;; 		  (apply sqlite3:map-row (lambda x x) sth data))
;;; 		 ((full-ping)  'full-ping)
;;; 		 (else (print "Not ready for action " action) #f)))
;;; 	      (loop (add1 count)
;;; 		    (if cookie
;;; 			(cons witem responses)
;;; 			responses))))))))
;;; 
;;; ;; do up to 400ms of processing on each queue
;;; ;; - the work-queue-processor will allow the max 1200ms of work to complete but it will flag as overloaded
;;; ;; 
;;; (define (process-db-queries acfg fname)
;;;   (if (hash-table-exists? (area-wqueues acfg) fname)
;;;       (let* ((process-db-queries-start-time (current-milliseconds))
;;; 	     (qdat             (hash-table-ref/default (area-wqueues acfg) fname #f))
;;; 	     (queue-sym->queue (lambda (queue-sym)
;;; 				 (case queue-sym  ;; lookup the queue from qdat given a name (symbol)
;;; 				   ((wqueue)  (qdat-writeq qdat))
;;; 				   ((rqueue)  (qdat-readq  qdat))
;;; 				   ((rwqueue) (qdat-rwq    qdat))
;;; 				   ((misc)    (qdat-misc   qdat))
;;; 				   (else #f))))
;;; 	     (dbdat   (get-dbh acfg fname))
;;; 	     (dbh     (if (dbdat? dbdat)(dbdat-dbh dbdat) #f))
;;; 	     (nowtime (current-seconds)))
;;; 	;; handle the queues that require a transaction
;;; 	;;
;;; 	(map ;; 
;;; 	 (lambda (queue-sym)
;;; 	   ;; (print "processing queue " queue-sym)
;;; 	   (let* ((queue (queue-sym->queue queue-sym)))
;;; 	     (if (not (queue-empty? queue))
;;; 		 (let ((responses
;;; 			(sqlite3:with-transaction ;; todo - catch exceptions...
;;; 			 dbh
;;; 			 (lambda ()
;;; 			   (let* ((res (doqueue acfg queue fname dbdat dbh))) ;; this does the work!
;;; 			     ;; (print "res=" res)
;;; 			     (match res
;;; 			      ((count delta responses)
;;; 			       (update-stats acfg fname queue-sym delta count)
;;; 			       (sdbg> "process-db-queries" "sqlite3-transaction" process-db-queries-start-time #f #f)
;;; 			       responses) ;; return responses
;;; 			      (else
;;; 			       (print "ERROR: bad return data from doqueue " res)))
;;; 			     )))))
;;; 		   ;; having completed the transaction, send the responses.
;;; 		   ;; (print "INFO: sending " (length responses) " responses.")
;;; 		   (let loop ((responses-left responses))
;;; 		     (cond
;;; 		      ((null? responses-left)  #t)
;;; 		      (else
;;; 		       (let* ((witem    (car responses-left))
;;; 			      (response (cdr responses-left)))  
;;; 			 (call-deliver-response acfg (witem-ripaddr witem)(witem-rport witem)
;;; 						(witem-cookie witem)(witem-result witem)))
;;; 		       (loop (cdr responses-left))))))
;;; 		 )))
;;; 	 '(wqueue rwqueue rqueue))
;;; 	
;;; 	;; handle misc queue
;;; 	;;
;;; 	;; (print "processing misc queue")
;;; 	(let ((queue (queue-sym->queue 'misc)))
;;; 	  (doqueue acfg queue fname dbdat dbh))
;;; 	;; ....
;;; 	(save-dbh acfg fname dbdat)
;;; 	#t ;; just to let the tests know we got here
;;; 	)
;;;       #f ;; nothing processed
;;;       ))
;;; 
;;; ;; run all queues in parallel per db but sequentially per queue for that db.
;;; ;;  - process the queues every 500 or so ms
;;; ;;  - allow for long running queries to continue but all other activities for that
;;; ;;    db will be blocked.
;;; ;;
;;; (define (work-queue-processor acfg)
;;;   (let* ((threads (make-hash-table))) ;; fname => thread
;;;     (let loop ((fnames      (hash-table-keys (area-wqueues acfg)))
;;; 	       (target-time (+ (current-milliseconds) 50)))
;;;       ;;(if (not (null? fnames))(print "Processing for these databases: " fnames))
;;;       (for-each
;;;        (lambda (fname)
;;; 	 ;; (print "processing for " fname)
;;; 	 ;;(process-db-queries acfg fname))
;;; 	 (let ((th (hash-table-ref/default threads fname #f)))
;;; 	   (if (and th (not (member (thread-state th) '(dead terminated))))
;;; 	       (begin
;;; 		 (print "WARNING: worker thread for " fname " is taking a long time.")
;;; 		 (print "Thread is in state " (thread-state th)))
;;; 	       (let ((th1 (make-thread (lambda ()
;;; 					 ;; (catch-and-dump
;;; 					 ;;  (lambda ()
;;; 					    ;; (print "Process queries for " fname)
;;; 					    (let ((start-time (current-milliseconds)))
;;; 					      (process-db-queries acfg fname)
;;; 					      ;; (thread-sleep! 0.01) ;; need the thread to take at least some time
;;; 					      (hash-table-delete! threads fname)) ;; no mutexes?
;;; 					    fname)
;;; 					  "th1"))) ;; ))
;;; 		 (hash-table-set! threads fname th1)
;;; 		 (thread-start! th1)))))
;;;        fnames)
;;;       ;; (thread-sleep! 0.1) ;; give the threads some time to process requests
;;;       ;; burn time until 400ms is up
;;;       (let ((now-time (current-milliseconds)))
;;; 	(if (< now-time target-time)
;;; 	    (let ((delta (- target-time now-time)))
;;; 	      (thread-sleep! (/ delta 1000)))))
;;;       (loop (hash-table-keys (area-wqueues acfg))
;;; 	    (+ (current-milliseconds) 50)))))
;;; 
;;; ;;======================================================================
;;; ;; S T A T S   G A T H E R I N G
;;; ;;======================================================================
;;; 
;;; (defstruct stat
;;;   (qcount-avg  0)                  ;; coarse running average
;;;   (qtime-avg   0)                  ;; coarse running average
;;;   (qcount      0)                  ;; total
;;;   (qtime       0)                  ;; total
;;;   (last-qcount 0)                  ;; last 
;;;   (last-qtime  0)                  ;; last
;;;   (dbs        '())                 ;; list of db files handled by this node
;;;   (when        0))                 ;; when the last query happened - seconds
;;; 
;;; 
;;; (define (update-stats acfg fname bucket duration numqueries)
;;;   (let* ((key   fname) ;; for now do not use bucket. Was: (conc fname "-" bucket)) ;; lazy but good enough
;;; 	 (stats (or (hash-table-ref/default (area-stats acfg) key #f)
;;; 		    (let ((newstats (make-stat)))
;;; 		      (hash-table-set! (area-stats acfg) key newstats)
;;; 		      newstats))))
;;;     ;; when the last query happended (used to remove the fname from the active list)
;;;     (stat-when-set! stats (current-seconds))
;;;     ;; last values
;;;     (stat-last-qcount-set! stats numqueries)
;;;     (stat-last-qtime-set!  stats duration)
;;;     ;; total over process lifetime
;;;     (stat-qcount-set! stats (+ (stat-qcount stats) numqueries))
;;;     (stat-qtime-set!  stats (+ (stat-qtime  stats) duration))
;;;     ;; coarse average
;;;     (stat-qcount-avg-set! stats (/ (+ (stat-qcount-avg stats) numqueries) 2))
;;;     (stat-qtime-avg-set!  stats (/ (+ (stat-qtime-avg  stats) duration)   2))
;;; 
;;;     ;; here is where we add the stats for a given dbfile
;;;     (if (not (member fname (stat-dbs stats)))
;;; 	(stat-dbs-set! stats (cons fname (stat-dbs stats))))
;;; 
;;;     ))
;;; 
;;; ;;======================================================================
;;; ;; S E R V E R   S T U F F 
;;; ;;======================================================================
;;; 
;;; ;; this does NOT return!
;;; ;;
;;; (define (find-free-port-and-open acfg)
;;;   (let ((port (or (area-port acfg) 3200)))
;;;     (handle-exceptions
;;; 	exn
;;; 	(begin
;;; 	  (print "INFO: cannot bind to port " (rpc:default-server-port) ", trying next port")
;;; 	  (area-port-set! acfg (+ port 1))
;;; 	  (find-free-port-and-open acfg))
;;;       (rpc:default-server-port port)
;;;       (area-port-set! acfg port)
;;;       (tcp-read-timeout 120000)
;;;       ;; ((rpc:make-server (tcp-listen port)) #t)
;;;       (tcp-listen (rpc:default-server-port)
;;;       ))))
;;; 
;;; ;; register this node by putting a packet into the pkts dir.
;;; ;; look for other servers
;;; ;; contact other servers and compile list of servers
;;; ;; there are two types of server
;;; ;;     main servers - dashboards, runners and dedicated servers - need pkt
;;; ;;     passive servers - test executers, step calls, list-runs - no pkt
;;; ;;
;;; (define (register-node acfg hostip port-num)
;;;   ;;(mutex-lock! (area-mutex acfg))
;;;   (let* ((server-type  (area-server-type acfg)) ;; auto, main, passive (no pkt created)
;;; 	 (best-ip      (or hostip (get-my-best-address)))
;;; 	 (mtdir        (area-dbdir acfg))
;;; 	 (pktdir       (area-pktsdir acfg))) ;; conc mtdir "/.server-pkts")))
;;;     (print "Registering node " best-ip ":" port-num)
;;;     (if (not mtdir) ;; require a home for this node to put or find databases
;;; 	#f
;;; 	(begin
;;; 	  (if  (not (directory? pktdir))(create-directory pktdir))
;;; 	  ;; server is started, now create pkt if needed
;;; 	  (print "Starting server in " server-type " mode with port " port-num)
;;; 	  (if (member server-type '(auto main)) ;; TODO: if auto, count number of servers registers, if > 3 then don't put out a pkt
;;; 	      (begin
;;; 		(area-pktid-set! acfg
;;; 				 (write-alist->pkt
;;; 				  pktdir 
;;; 				  `((hostname . ,(get-host-name))
;;; 				    (ipaddr   . ,best-ip)
;;; 				    (port     . ,port-num)
;;; 				    (pid      . ,(current-process-id)))
;;; 				  pktspec: *pktspec*
;;; 				  ptype:   'server))
;;; 		(area-pktfile-set! acfg (conc pktdir "/" (area-pktid acfg) ".pkt"))))
;;; 	  (area-port-set!    acfg port-num)
;;; 	  #;(mutex-unlock! (area-mutex acfg))))))
;;; 
;;; (define *cookie-seqnum* 0)
;;; (define (make-cookie key)
;;;   (set! *cookie-seqnum* (add1 *cookie-seqnum*))
;;;   ;;(print "MAKE COOKIE CALLED -- on "servkey"-"*cookie-seqnum*)
;;;   (conc key "-" *cookie-seqnum*)
;;;   )
;;; 
;;; ;; dispatch locally if possible
;;; ;;
;;; (define (call-deliver-response acfg ipaddr port cookie data)
;;;   (if (and (equal? (area-myaddr acfg) ipaddr)
;;; 	   (equal? (area-port     acfg) port))
;;;       (deliver-response acfg cookie data)
;;;       ((rpc:procedure 'response ipaddr port) cookie data)))
;;; 
;;; (define (deliver-response acfg cookie data)
;;;   (let ((deliver-response-start (current-milliseconds)))
;;;     (thread-start! (make-thread
;;; 		    (lambda ()
;;; 		      (let loop ((tries-left 5))
;;; 			;;(print "TOP OF DELIVER_RESPONSE LOOP; triesleft="tries-left)
;;; 			;;(pp (hash-table->alist (area-cookie2mbox acfg)))
;;; 			(let* ((mbox (hash-table-ref/default (area-cookie2mbox acfg) cookie #f)))
;;; 			  (cond
;;; 			   ((eq? 0 tries-left)
;;; 			    (print "ulex:deliver-response: I give up. Mailbox never appeared. cookie="cookie)
;;; 			    )
;;; 			   (mbox
;;; 			    ;;(print "got mbox="mbox"  got data="data"  send.")
;;; 			    (mailbox-send! mbox data))
;;; 			   (else
;;; 			    ;;(print "no mbox yet.  look for "cookie)
;;; 			    (thread-sleep! (/ (- 6 tries-left) 10))
;;; 			    (loop (sub1 tries-left))))))
;;; 		      ;; (debug-pp (list (conc "ulex:deliver-response took " (- (current-milliseconds) deliver-response-start) " ms, cookie=" cookie " data=") data))
;;; 		      (sdbg> "deliver-response" "mailbox-send" deliver-response-start #f #f cookie)
;;; 		      )
;;; 		    (conc "deliver-response thread for cookie="cookie))))
;;;   #t)
;;; 
;;; ;; action:
;;; ;;   immediate - quick actions, no need to put in queues
;;; ;;   dbwrite   - put in dbwrite queue
;;; ;;   dbread    - put in dbread queue
;;; ;;   oslong    - os actions, e.g. du, that could take a long time
;;; ;;   osshort   - os actions that should be quick, e.g. df
;;; ;;
;;; (define (request acfg from-ipaddr from-port servkey action cookie fname params) ;; std-peer-handler
;;;   ;; NOTE: Use rpc:current-peer for getting return address
;;;   (let* ((std-peer-handler-start (current-milliseconds))
;;; 	 ;; (raw-data               (alist-ref 'data     dat))
;;; 	 (rdat                   (hash-table-ref/default
;;; 				  (area-rtable acfg) action #f)) ;; this looks up the sql query or other details indexed by the action
;;; 	 (witem                  (make-witem ripaddr: from-ipaddr ;; rhost:   from-host   
;;; 					     rport:   from-port   action:  action
;;; 					     rdat:    rdat        cookie:  cookie
;;; 					     servkey: servkey     data:    params ;; TODO - rename data to params
;;; 					     caller:  (rpc:current-peer))))
;;;     (if (not (equal? servkey (area-pktid acfg)))
;;; 	`(#f . ,(conc "I don't know you servkey=" servkey ", pktid=" (area-pktid acfg))) ;; immediately return this
;;; 	(let* ((ctype (if rdat 
;;; 			  (calldat-ctype rdat) ;; is this necessary? these should be identical
;;; 			  action)))
;;; 	  (sdbg> "std-peer-handler" "immediate" std-peer-handler-start #f #f)
;;; 	  (case ctype
;;; 	    ;; (dbwrite acfg rdat (cons from-ipaddr from-port) data)))
;;; 	    ((full-ping)  `(#t  "ack to full ping"        ,(work-queue-add acfg fname witem) ,cookie))
;;; 	    ((response)   `(#t  "ack from requestor"      ,(deliver-response acfg fname params)))
;;; 	    ((dbwrite)    `(#t  "db write submitted"      ,(work-queue-add acfg fname witem) ,cookie))
;;; 	    ((dbread)     `(#t  "db read submitted"       ,(work-queue-add acfg fname witem) ,cookie  ))
;;; 	    ((dbrw)       `(#t  "db read/write submitted" ,cookie))
;;; 	    ((osshort)    `(#t  "os short submitted"      ,cookie))
;;; 	    ((oslong)     `(#t  "os long submitted"       ,cookie))
;;; 	    (else         `(#f  "unrecognised action"     ,ctype)))))))
;;; 
;;; ;; Call this to start the actual server
;;; ;;
;;; ;; start_server
;;; ;;
;;; ;;   mode: '
;;; ;;   handler: proc which takes pktrecieved as argument
;;; ;;
;;; 
;;; (define (start-server acfg)
;;;   (let* ((conn (find-free-port-and-open acfg))
;;; 	 (port (area-port acfg)))
;;;     (rpc:publish-procedure!
;;;      'delist-db
;;;      (lambda (fname)
;;;        (hash-table-delete! (area-dbs acfg) fname)))
;;;     (rpc:publish-procedure!
;;;      'calling-addr
;;;      (lambda ()
;;;        (rpc:current-peer)))
;;;     (rpc:publish-procedure!
;;;      'ping
;;;      (lambda ()(real-ping acfg)))
;;;     (rpc:publish-procedure!
;;;      'request
;;;      (lambda (from-addr from-port servkey action cookie dbname params)
;;;        (request acfg from-addr from-port servkey action cookie dbname params)))
;;;     (rpc:publish-procedure!
;;;      'response
;;;      (lambda (cookie res-dat)
;;;        (deliver-response acfg cookie res-dat)))
;;;     (area-ready-set! acfg #t)
;;;     (area-conn-set! acfg conn)
;;;     ((rpc:make-server conn) #f)));; ((tcp-listen (rpc:default-server-port)) #t)
;;; 
;;; 
;;; (define (launch acfg) ;;  #!optional (proc std-peer-handler))
;;;   (print "starting launch")
;;;   (update-known-servers acfg) ;; gotta do this on every start (thus why limit number of publicised servers)
;;;   #;(let ((original-handler (current-exception-handler))) ;; is th
;;;     (lambda (exception)
;;;       (server-exit-procedure)
;;;       (original-handler exception)))
;;;   (on-exit (lambda ()
;;; 	     (shutdown acfg))) ;; (finalize-all-db-handles acfg)))
;;;   ;; set up the rpc handler
;;;   (let* ((th1  (make-thread
;;; 		(lambda ()(start-server acfg))
;;; 		"server thread"))
;;; 	 (th2   (make-thread
;;; 		 (lambda ()
;;; 		   (print "th2 starting")
;;; 		   (let loop ()
;;; 		     (work-queue-processor acfg)
;;; 		     (print "work-queue-processor crashed!")
;;; 		     (loop)))
;;; 		 "work queue thread")))
;;;     (thread-start! th1)
;;;     (thread-start! th2)
;;;     (let loop ()
;;;       (thread-sleep! 0.025)
;;;       (if (area-ready acfg)
;;; 	  #t
;;; 	  (loop)))
;;;     ;; attempt to fix my address
;;;     (let* ((all-addr (get-all-ips-sorted)))	     ;; could use (tcp-addresses conn)?
;;;       (let loop ((rem-addrs all-addr))
;;; 	(if (null? rem-addrs)
;;; 	    (begin
;;; 	      (print "ERROR: Failed to figure out the ip address of myself as a server. Giving up.")
;;; 	      (exit 1)) ;; BUG Changeme to raising an exception
;;; 		
;;; 	    (let* ((addr      (car rem-addrs))
;;; 		   (good-addr (handle-exceptions
;;; 				  exn
;;; 				  #f
;;; 				((rpc:procedure 'calling-addr addr (area-port acfg))))))
;;; 	      (if good-addr
;;; 		  (begin
;;; 		    (print "Got good-addr of " good-addr)
;;; 		    (area-myaddr-set! acfg good-addr))
;;; 		  (loop (cdr rem-addrs)))))))
;;;     (register-node acfg (area-myaddr acfg)(area-port acfg))
;;;     (print "INFO: Server started on " (area-myaddr acfg) ":" (area-port acfg))
;;;     ;; (update-known-servers acfg) ;; gotta do this on every start (thus why limit number of publicised servers)
;;;     ))
;;; 
;;; (define (clear-server-pkt acfg)
;;;   (let ((pktf (area-pktfile acfg)))
;;;     (if pktf (delete-file* pktf))))
;;; 
;;; (define (shutdown acfg)
;;;   (let (;;(conn (area-conn    acfg))
;;; 	(pktf (area-pktfile acfg))
;;; 	(port (area-port    acfg)))
;;;     (if pktf (delete-file* pktf))
;;;     (send-all "imshuttingdown")
;;;     ;; (rpc:close-all-connections!) ;; don't know if this is actually needed
;;;     (finalize-all-db-handles acfg)))
;;; 
;;; (define (send-all msg)
;;;   #f)
;;; 
;;; ;; given a area record look up all the packets
;;; ;;
;;; (define (get-all-server-pkts acfg)
;;;   (let ((all-pkt-files (glob (conc (area-pktsdir acfg) "/*.pkt"))))
;;;     (map (lambda (pkt-file)
;;; 	   (read-pkt->alist pkt-file pktspec: *pktspec*))
;;; 	 all-pkt-files)))
;;; 
;;; #;((Z . "9a0212302295a19610d5796fce0370fa130758e9")
;;;   (port . "34827")
;;;   (pid . "28748")
;;;   (hostname . "zeus")
;;;   (T . "server")
;;;   (D . "1549427032.0"))
;;; 
;;; #;(define (get-my-best-address)
;;;   (let ((all-my-addresses (get-all-ips))) ;; (vector->list (hostinfo-addresses (hostname->hostinfo (get-host-name))))))
;;;     (cond
;;;      ((null? all-my-addresses)
;;;       (get-host-name))                                          ;; no interfaces?
;;;      ((eq? (length all-my-addresses) 1)
;;;       (ip->string (car all-my-addresses)))                      ;; only one to choose from, just go with it
;;;      (else 
;;;       (ip->string (car (filter (lambda (x)                      ;; take any but 127.
;;; 				 (not (eq? (u8vector-ref x 0) 127)))
;;; 			       all-my-addresses)))))))
;;; 
;;; ;; whoami? I am my pkt
;;; ;;
;;; (define (whoami? acfg)
;;;   (hash-table-ref/default (area-hosts acfg)(area-pktid acfg) #f))
;;; 
;;; ;;======================================================================
;;; ;; "Client side" operations
;;; ;;======================================================================
;;; 
;;; (define (safe-call call-key host port . params)
;;;   (handle-exceptions
;;;    exn
;;;    (begin
;;;      (print "Call " call-key " to " host ":" port " failed")
;;;      #f)
;;;    (apply (rpc:procedure call-key host port) params)))
;;; 
;;; ;; ;; convert to/from string / sexpr
;;; ;; 
;;; ;; (define (string->sexpr str)
;;; ;;   (if (string? str)
;;; ;;       (with-input-from-string str read)
;;; ;;       str))
;;; ;; 
;;; ;; (define (sexpr->string s)
;;; ;;   (with-output-to-string (lambda ()(write s))))
;;; 
;;; ;; is the server alive?
;;; ;;
;;; (define (ping acfg host port)
;;;   (let* ((myaddr     (area-myaddr acfg))
;;; 	 (myport     (area-port   acfg))
;;; 	 (start-time (current-milliseconds))
;;; 	 (res        (if (and (equal? myaddr host)
;;; 			      (equal? myport port))
;;; 			 (real-ping acfg)
;;; 			 ((rpc:procedure 'ping host port)))))
;;;     (cons (- (current-milliseconds) start-time)
;;; 	  res)))
;;; 
;;; ;; returns ( ipaddr port alist-fname=>randnum )
;;; (define (real-ping acfg)
;;;   `(,(area-myaddr acfg) ,(area-port acfg) ,(get-host-stats acfg)))
;;; 
;;; ;; is the server alive AND the queues processing?
;;; ;;
;;; #;(define (full-ping acfg servpkt)
;;;   (let* ((start-time (current-milliseconds))
;;; 	 (res        (send-message acfg servpkt '(full-ping) 'full-ping)))
;;;     (cons (- (current-milliseconds) start-time)
;;; 	  res))) ;; (equal? res "got ping"))))
;;; 
;;; 
;;; ;; look up all pkts and get the server id (the hash), port, host/ip
;;; ;; store this info in acfg
;;; ;; return the number of responsive servers found
;;; ;;
;;; ;; DO NOT VERIFY THAT THE SERVER IS ALIVE HERE. This is called at times where the current server is not yet alive and cannot ping itself
;;; ;;
;;; (define (update-known-servers acfg)
;;;   ;; readll all pkts
;;;   ;; foreach pkt; if it isn't me ping the server; if alive, add to hosts hash, else rm the pkt
;;;   (let* ((start-time (current-milliseconds))
;;; 	 (all-pkts  (delete-duplicates
;;; 		     (append (get-all-server-pkts acfg)
;;; 			     (hash-table-values (area-hosts acfg)))))
;;; 	 (hostshash (area-hosts acfg))
;;; 	 (my-id     (area-pktid acfg))
;;; 	 (pktsdir   (area-pktsdir acfg)) ;; needed to remove pkts from non-responsive servers
;;; 	 (numsrvs   0)
;;; 	 (delpkt    (lambda (pktsdir sid)
;;; 		      (print "clearing out server " sid)
;;; 		      (delete-file* (conc pktsdir "/" sid ".pkt"))
;;; 		      (hash-table-delete! hostshash sid))))
;;;     (area-last-srvup-set! acfg (current-seconds))
;;;     (for-each
;;;      (lambda (servpkt)
;;;        (if (list? servpkt)
;;; 	   ;; (pp servpkt)
;;; 	   (let* ((shost (alist-ref 'ipaddr servpkt))
;;; 		  (sport (any->number (alist-ref 'port servpkt)))
;;; 		  (res   (handle-exceptions
;;; 			  exn
;;; 			  (begin
;;; 			    ;; (print "INFO: bad server on " shost ":" sport)
;;; 			    #f)
;;; 			  (ping acfg shost sport)))
;;; 		  (sid   (alist-ref 'Z servpkt)) ;; Z code is our name for the server
;;; 		  (url   (conc shost ":" sport))
;;; 		  )
;;; 	     #;(if (or (not res)
;;; 		     (null? res))
;;; 		 (begin
;;; 		   (print "STRANGE: ping of " url " gave " res)))
;;; 	     
;;; 	     ;; (print "Got " res " from " shost ":" sport)
;;; 	     (match res
;;; 		    ((qduration . payload)
;;; 		     ;; (print "Server pkt:" (alist-ref 'ipaddr servpkt) ":" (alist-ref 'port servpkt)
;;; 		     ;;        (if payload
;;; 		     ;;            "Success" "Fail"))
;;; 		     (match payload
;;; 			    ((host port stats)
;;; 			     ;; (print "From " host ":" port " got stats: " stats)
;;; 			     (if (and host port stats)
;;; 				 (let ((url (conc host ":" port)))
;;; 				   (hash-table-set! hostshash sid servpkt)
;;; 				   ;; store based on host:port
;;; 				   (hash-table-set! (area-hoststats acfg) sid stats))
;;; 				 (print "missing data from the server, not sure what that means!"))
;;; 			     (set! numsrvs (+ numsrvs 1)))
;;; 			    (#f
;;; 			     (print "Removing pkt " sid " due to #f from server or failed ping")
;;; 			     (delpkt pktsdir sid))
;;; 			    (else
;;; 			     (print "Got ")(pp res)(print " from server ")(pp servpkt) " but response did not match (#f/#t . msg)")))
;;; 		    (else
;;; 		     ;; here we delete the pkt - can't reach the server, remove it
;;; 		     ;; however this logic is inadequate. we should mark the server as checked
;;; 		     ;; and not good, if it happens a second time - then remove the pkt
;;; 		     ;; or something similar. I.e. don't be too quick to assume the server is wedged or dead
;;; 		     ;; could be it is simply too busy to reply
;;; 		     (let ((bad-pings (hash-table-ref/default (area-health acfg) url 0)))
;;; 		       (if (> bad-pings 1) ;; two bad pings - remove pkt
;;; 			   (begin
;;; 			     (print "INFO: " bad-pings " bad responses from " url ", deleting pkt " sid)
;;; 			     (delpkt pktsdir sid))
;;; 			   (begin
;;; 			     (print "INFO: " bad-pings " bad responses from " shost ":" sport " not deleting pkt yet")
;;; 			     (hash-table-set! (area-health acfg)
;;; 					      url
;;; 					      (+ (hash-table-ref/default (area-health acfg) url 0) 1))
;;; 			     ))
;;; 		       ))))
;;; 	   ;; servpkt is not actually a pkt?
;;; 	   (begin
;;; 	     (print "Bad pkt " servpkt))))
;;;      all-pkts)
;;;     (sdbg> "update-known-servers" "end" start-time #f #f " found " numsrvs
;;; 	   " servers, pkts: " (map (lambda (p)
;;; 				     (alist-ref 'Z p))
;;; 				   all-pkts))
;;;     numsrvs))
;;; 
;;; (defstruct srvstat
;;;   (numfiles 0)   ;; number of db files handled by this server - subtract 1 for the db being currently looked at
;;;   (randnum  #f)  ;; tie breaker number assigned to by the server itself - applies only to the db under consideration
;;;   (pkt      #f)) ;; the server pkt
;;; 
;;; ;;(define (srv->srvstat srvpkt)
;;;   
;;; ;; Get the server best for given dbname and key
;;; ;;
;;; ;;   NOTE: key is not currently used. The key points to the kind of query, this may be useful for directing read-only queries.
;;; ;;
;;; (define (get-best-server acfg dbname key)
;;;   (let* (;; (servers (hash-table-values (area-hosts acfg)))
;;; 	 (servers     (area-hosts acfg))
;;; 	 (skeys       (sort (hash-table-keys servers) string>=?)) ;; a stable listing
;;; 	 (start-time  (current-milliseconds))
;;; 	 (srvstats    (make-hash-table))  ;; srvid => srvstat
;;; 	 (url         (conc (area-myaddr acfg) ":" (area-port acfg))))
;;;     ;; (print "scores for " dbname ": " (map (lambda (k)(cons k (calc-server-score acfg dbname k))) skeys))
;;;     (if (null? skeys)
;;; 	(if (> (update-known-servers acfg) 0)
;;; 	    (get-best-server acfg dbname key) ;; some risk of infinite loop here, TODO add try counter
;;; 	    (begin
;;; 	      (print "ERROR: no server found!") ;; since this process is also a server this should never happen
;;; 	      #f))
;;; 	(begin
;;; 	  ;; (print "in get-best-server with skeys=" skeys)
;;; 	  (if (> (- (current-seconds) (area-last-srvup acfg)) 10)
;;; 	      (begin
;;; 		(update-known-servers acfg)
;;; 		(sdbg> "get-best-server" "update-known-servers" start-time #f #f)))
;;; 
;;; 	  ;; for each server look at the list of dbfiles, total number of dbs being handled
;;; 	  ;; and the rand number, save the best host
;;; 	  ;; also do a delist-db for each server dbfile not used
;;; 	  (let* ((best-server       #f)
;;; 		 (servers-to-delist (make-hash-table)))
;;; 	    (for-each
;;; 	     (lambda (srvid)
;;; 	       (let* ((server    (hash-table-ref/default servers srvid #f))
;;; 		      (stats     (hash-table-ref/default (area-hoststats acfg) srvid '(()))))
;;; 		 ;; (print "stats: " stats)
;;;  		 (if server
;;; 		     (let* ((dbweights (car stats))
;;; 			    (srvload   (length (filter (lambda (x)(not (equal? dbname (car x)))) dbweights)))
;;; 			    (dbrec     (alist-ref dbname dbweights equal?))  ;; get the pair with fname . randscore
;;; 			    (randnum   (if dbrec
;;; 					   dbrec ;; (cdr dbrec)
;;; 					   0)))
;;; 		       (hash-table-set! srvstats srvid (make-srvstat numfiles: srvload randnum: randnum pkt: server))))))
;;; 	     skeys)
;;; 	    
;;; 	    (let* ((sorted    (sort (hash-table-values srvstats) 
;;; 				    (lambda (a b)
;;; 				      (let ((numfiles-a (srvstat-numfiles a))
;;; 					    (numfiles-b (srvstat-numfiles b))
;;; 					    (randnum-a  (srvstat-randnum a))
;;; 					    (randnum-b  (srvstat-randnum b)))
;;; 					(if (< numfiles-a numfiles-b) ;; Note, I don't think adding an offset works here. Goal was only move file handling to a different server if it has 2 less
;;; 					    #t
;;; 					    (if (and (equal? numfiles-a numfiles-b)
;;; 						     (< randnum-a randnum-b))
;;; 						#t
;;; 						#f))))))
;;; 		   (best      (if (null? sorted)
;;; 				  (begin
;;; 				    (print "ERROR: should never be null due to self as server.")
;;; 				    #f)
;;; 				  (srvstat-pkt (car sorted)))))
;;; 	      #;(print "SERVER(" url "): " dbname ": " (map (lambda (srv)
;;; 							    (let ((p (srvstat-pkt srv)))
;;; 							      (conc (alist-ref 'ipaddr p) ":" (alist-ref 'port p)
;;; 								    "(" (srvstat-numfiles srv)","(srvstat-randnum srv)")")))
;;; 							    sorted))
;;; 	      best))))))
;;;     
;;;     ;; send out an "I'm about to exit notice to all known servers"
;;;     ;;
;;; (define (death-imminent acfg)
;;;   '())
;;; 
;;; ;;======================================================================
;;; ;; U L E X  -  T H E   I N T E R E S T I N G   S T U F F ! !
;;; ;;======================================================================
;;; 
;;; ;; register a handler
;;; ;;   NOTES:
;;; ;;     dbinitsql   is reserved for a list of sql statements for initializing the db
;;; ;;     dbinitfn    is reserved for a db init function, if exists called after dbinitsql
;;; ;;     
;;; (define (register acfg key obj #!optional (ctype 'dbwrite))
;;;   (let ((ht (area-rtable acfg)))
;;;     (if (hash-table-exists? ht key)
;;; 	(print "WARNING: redefinition of entry " key))
;;;     (hash-table-set! ht key (make-calldat obj: obj ctype: ctype))))
;;; 
;;; ;; usage: register-batch acfg '((key1 . sql1) (key2 . sql2) ... )
;;; ;; NB// obj is often an sql query
;;; ;;
;;; (define (register-batch acfg ctype data)
;;;   (let ((ht (area-rtable acfg)))
;;;     (map (lambda (dat)
;;; 	   (hash-table-set! ht (car dat)(make-calldat obj: (cdr dat) ctype: ctype)))
;;; 	 data)))
;;; 
;;; (define (initialize-area-calls-from-specfile area specfile)
;;;   (let* ((callspec (with-input-from-file specfile read )))
;;;     (for-each (lambda (group)
;;;                 (register-batch
;;;                  area
;;;                  (car group)
;;;                  (cdr group)))
;;;               callspec)))
;;; 
;;; ;; get-rentry
;;; ;;
;;; (define (get-rentry acfg key)
;;;   (hash-table-ref/default (area-rtable acfg) key #f))
;;; 
;;; (define (get-rsql acfg key)
;;;   (let ((cdat (get-rentry acfg key)))
;;;     (if cdat
;;; 	(calldat-obj cdat)
;;; 	#f)))
;;; 
;;; 
;;; 
;;; ;; blocking call:
;;; ;;    client                         server
;;; ;;    ------                         ------
;;; ;;    call()
;;; ;;    send-message()
;;; ;;    nmsg-send()
;;; ;;                                   nmsg-receive()
;;; ;;                                   nmsg-respond(ack,cookie)
;;; ;;    ack, cookie
;;; ;;    mbox-thread-wait(cookie)
;;; ;;                                   nmsg-send(client,cookie,result)
;;; ;;        nmsg-respond(ack)
;;; ;;        return result
;;; ;;
;;; ;; reserved action:
;;; ;;    'immediate
;;; ;;    'dbinitsql
;;; ;;
;;; (define (call acfg dbname action params #!optional (count 0))
;;;   (let* ((call-start-time     (current-milliseconds))
;;; 	 (srv                 (get-best-server acfg dbname action))
;;; 	 (post-get-start-time (current-milliseconds))
;;; 	 (rdat                (hash-table-ref/default (area-rtable acfg) action #f))
;;; 	 (myid                (trim-pktid (area-pktid acfg)))
;;; 	 (srvid               (trim-pktid (alist-ref 'Z srv)))
;;; 	 (cookie              (make-cookie myid)))
;;;     (sdbg> "call" "get-best-server" call-start-time #f call-start-time " from: " myid " to server: " srvid " for " dbname " action: " action " params: " params " rdat: " rdat)
;;;     (print "INFO: call to " (alist-ref 'ipaddr srv) ":" (alist-ref 'port srv) " from " (area-myaddr acfg) ":" (area-port acfg) " for " dbname)
;;;     (if (and srv rdat) ;; need both to dispatch a request
;;; 	(let* ((ripaddr  (alist-ref 'ipaddr srv))
;;; 	       (rsrvid   (alist-ref 'Z srv))
;;; 	       (rport    (any->number (alist-ref 'port   srv)))
;;; 	       (res-full (if (and (equal? ripaddr (area-myaddr acfg))
;;; 				  (equal? rport   (area-port acfg)))
;;; 			     (request acfg ripaddr rport (area-pktid acfg) action cookie dbname params)
;;; 			     (safe-call 'request ripaddr rport
;;; 					(area-myaddr acfg)
;;; 					(area-port   acfg)
;;; 					#;(area-pktid acfg)
;;; 					rsrvid
;;; 					action cookie dbname params))))
;;; 	  ;; (print "res-full: " res-full)
;;; 	  (match res-full
;;; 	    ((response-ok response-msg rem ...)
;;; 	     (let* ((send-message-time (current-milliseconds))
;;; 		    ;; (match res-full
;;; 		    ;;  ((response-ok response-msg)
;;; 		    ;; (response-ok  (car res-full))
;;; 		    ;; (response-msg (cadr res-full)
;;; 		    )
;;; 	       ;; (res (take res-full 3))) ;; ctype == action, TODO: converge on one term <<=== what was this? BUG 
;;; 	       ;; (print "ulex:call: send-message took " (- send-message-time post-get-start-time) " ms params=" params)
;;; 	       (sdbg> "call" "send-message" post-get-start-time #f call-start-time)
;;; 	       (cond
;;; 		((not response-ok) #f)
;;; 		((member response-msg '("db read submitted" "db write submitted"))
;;; 		 (let* ((cookie-id   (cadddr res-full))
;;; 			(mbox        (make-mailbox))
;;; 			(mbox-time   (current-milliseconds)))
;;; 		   (hash-table-set! (area-cookie2mbox acfg) cookie-id mbox)
;;; 		   (let* ((mbox-timeout-secs    20)
;;; 			  (mbox-timeout-result 'MBOX_TIMEOUT)
;;; 			  (res                  (mailbox-receive! mbox mbox-timeout-secs mbox-timeout-result))
;;; 			  (mbox-receive-time    (current-milliseconds)))
;;; 		     (hash-table-delete! (area-cookie2mbox acfg) cookie-id)
;;; 		     (sdbg> "call" "mailbox-receive" mbox-time #f call-start-time " from: " myid " to server: " srvid " for " dbname)
;;; 		     ;; (print "ulex:call mailbox-receive took " (- mbox-receive-time mbox-time) "ms params=" params)
;;; 		     res)))
;;; 		(else
;;; 		 (print "Unhandled response \""response-msg"\"")
;;; 		 #f))
;;; 	       ;; depending on what action (i.e. ctype) is we will block here waiting for
;;; 	       ;; all the data (mechanism to be determined)
;;; 	       ;;
;;; 	       ;; if res is a "working on it" then wait
;;; 	       ;;    wait for result
;;; 	       ;; mailbox thread wait on 
;;; 	       
;;; 	       ;; if res is a "can't help you" then try a different server
;;; 	       ;; if res is a "ack" (e.g. for one-shot requests) then return res
;;; 	       ))
;;; 	    (else
;;; 	     (if (< count 10)
;;; 		 (let* ((url (conc (alist-ref 'ipaddr srv) ":" (alist-ref 'port srv))))
;;; 		   (thread-sleep! 1)
;;; 		   (print "ERROR: Bad result from " url ", dbname: " dbname ", action: " action ", params: " params ". Trying again in 1 second.")
;;; 		   (call acfg dbname action params (+ count 1)))
;;; 		 (begin
;;; 		   (error (conc "ERROR: " count " tries, still have improper response res-full=" res-full)))))))
;;; 	(begin
;;; 	  (if (not rdat)
;;; 	      (print "ERROR: action " action " not registered.")
;;; 	      (if (< count 10)
;;; 		 (begin
;;; 		   (thread-sleep! 1)
;;; 		   (area-hosts-set! acfg (make-hash-table)) ;; clear out all known hosts
;;; 		   (print "ERROR: no server found, srv=" srv ", trying again in 1 seconds")
;;; 		   (call acfg dbname action params (+ count 1)))
;;; 		 (begin
;;; 		   (error (conc "ERROR: no server found after 10 tries, srv=" srv ", giving up."))
;;; 		   #;(error "No server available"))))))))
;;; 
;;; 
;;; ;;======================================================================
;;; ;; U T I L I T I E S 
;;; ;;======================================================================
;;; 
;;; ;; get a signature for identifing this process
;;; ;;
;;; (define (get-process-signature)
;;;   (cons (get-host-name)(current-process-id)))
;;; 
;;; ;;======================================================================
;;; ;; S Y S T E M   S T U F F
;;; ;;======================================================================
;;; 
;;; ;; get normalized cpu load by reading from /proc/loadavg and
;;; ;; /proc/cpuinfo return all three values and the number of real cpus
;;; ;; and the number of threads returns alist '((adj-cpu-load
;;; ;; . normalized-proc-load) ... etc.  keys: adj-proc-load,
;;; ;; adj-core-load, 1m-load, 5m-load, 15m-load
;;; ;;
;;; (define (get-normalized-cpu-load)
;;;   (let ((res (get-normalized-cpu-load-raw))
;;; 	(default `((adj-proc-load . 2) ;; there is no right answer
;;; 		   (adj-core-load . 2)
;;; 		   (1m-load       . 2)
;;; 		   (5m-load       . 0) ;; causes a large delta - thus causing default of throttling if stuff goes wrong
;;; 		   (15m-load      . 0)
;;; 		   (proc          . 1)
;;; 		   (core          . 1)
;;; 		   (phys          . 1)
;;; 		   (error         . #t))))
;;;     (cond
;;;      ((and (list? res)
;;; 	   (> (length res) 2))
;;;       res)
;;;      ((eq? res #f)   default) ;; add messages?
;;;      ((eq? res #f) default)   ;; this would be the #eof
;;;      (else default))))
;;; 
;;; (define (get-normalized-cpu-load-raw)
;;;   (let* ((actual-host           (get-host-name))) ;; #f is localhost
;;;     (let ((data  (append 
;;; 		  (with-input-from-file "/proc/loadavg" read-lines)
;;; 		  (with-input-from-file "/proc/cpuinfo" read-lines)
;;; 		  (list "end")))
;;; 	  (load-rx  (regexp "^([\\d\\.]+)\\s+([\\d\\.]+)\\s+([\\d\\.]+)\\s+.*$"))
;;; 	  (proc-rx  (regexp "^processor\\s+:\\s+(\\d+)\\s*$"))
;;; 	  (core-rx  (regexp "^core id\\s+:\\s+(\\d+)\\s*$"))
;;; 	  (phys-rx  (regexp "^physical id\\s+:\\s+(\\d+)\\s*$"))
;;; 	  (max-num  (lambda (p n)(max (string->number p) n))))
;;;       ;; (print "data=" data)
;;;       (if (null? data) ;; something went wrong
;;; 	  #f
;;; 	  (let loop ((hed      (car data))
;;; 		     (tal      (cdr data))
;;; 		     (loads    #f)
;;; 		     (proc-num 0)  ;; processor includes threads
;;; 		     (phys-num 0)  ;; physical chip on motherboard
;;; 		     (core-num 0)) ;; core
;;; 	    ;; (print hed ", " loads ", " proc-num ", " phys-num ", " core-num)
;;; 	    (if (null? tal) ;; have all our data, calculate normalized load and return result
;;; 		(let* ((act-proc (+ proc-num 1))
;;; 		       (act-phys (+ phys-num 1))
;;; 		       (act-core (+ core-num 1))
;;; 		       (adj-proc-load (/ (car loads) act-proc))
;;; 		       (adj-core-load (/ (car loads) act-core))
;;; 		       (result
;;; 			(append (list (cons 'adj-proc-load adj-proc-load)
;;; 				      (cons 'adj-core-load adj-core-load))
;;; 				(list (cons '1m-load (car loads))
;;; 				      (cons '5m-load (cadr loads))
;;; 				      (cons '15m-load (caddr loads)))
;;; 				(list (cons 'proc act-proc)
;;; 				      (cons 'core act-core)
;;; 				      (cons 'phys act-phys)))))
;;; 		  result)
;;; 		(regex-case
;;; 		    hed
;;; 		  (load-rx  ( x l1 l5 l15 ) (loop (car tal)(cdr tal)(map string->number (list l1 l5 l15)) proc-num phys-num core-num))
;;; 		  (proc-rx  ( x p         ) (loop (car tal)(cdr tal) loads           (max-num p proc-num) phys-num core-num))
;;; 		  (phys-rx  ( x p         ) (loop (car tal)(cdr tal) loads           proc-num (max-num p phys-num) core-num))
;;; 		  (core-rx  ( x c         ) (loop (car tal)(cdr tal) loads           proc-num phys-num (max-num c core-num)))
;;; 		  (else 
;;; 		   (begin
;;; 		     ;; (print "NO MATCH: " hed)
;;; 		     (loop (car tal)(cdr tal) loads proc-num phys-num core-num))))))))))
;;; 
;;; (define (get-host-stats acfg)
;;;   (let ((stats-hash (area-stats acfg)))
;;;     ;; use this opportunity to remove references to dbfiles which have not been accessed in a while
;;;     (for-each
;;;      (lambda (dbname)
;;;        (let* ((stats       (hash-table-ref stats-hash dbname))
;;; 	      (last-access (stat-when stats)))
;;; 	 (if (and (> last-access 0)                             ;; if zero then there has been no access
;;; 		  (> (- (current-seconds) last-access) 10))     ;; not used in ten seconds
;;; 	     (begin
;;; 	       (print "Removing " dbname " from stats list")
;;; 	       (hash-table-delete! stats-hash dbname) ;; remove from stats hash
;;; 	       (stat-dbs-set! stats (hash-table-keys stats))))))
;;;      (hash-table-keys stats-hash))
;;;     
;;;     `(,(hash-table->alist (area-dbs acfg)) ;; dbname => randnum
;;;       ,(map (lambda (dbname)  ;; dbname is the db name
;;; 	      (cons dbname (stat-when (hash-table-ref stats-hash dbname))))
;;; 	    (hash-table-keys stats-hash))
;;;       (cpuload . ,(get-normalized-cpu-load)))))
;;;     #;(stats   . ,(map (lambda (k) ;; create an alist from the stats data
;;; 		       (cons k (stat->alist (hash-table-ref (area-stats acfg) k))))
;;; 		     (hash-table-keys (area-stats acfg))))
;;; 
;;; #;(trace
;;;  ;; assv
;;;  ;; cdr
;;;  ;; caar
;;;  ;; ;; cdr
;;;  ;; call
;;;  ;; finalize-all-db-handles
;;;  ;; get-all-server-pkts
;;;  ;; get-normalized-cpu-load
;;;  ;; get-normalized-cpu-load-raw
;;;  ;; launch
;;;  ;; nmsg-send
;;;  ;; process-db-queries
;;;  ;; receive-message
;;;  ;; std-peer-handler
;;;  ;; update-known-servers
;;;  ;; work-queue-processor
;;;  )
;;; 
;;; ;;======================================================================
;;; ;; netutil
;;; ;;   move this back to ulex-netutil.scm someday?
;;; ;;======================================================================
;;; 
;;; ;; #include <stdio.h>
;;; ;; #include <netinet/in.h>
;;; ;; #include <string.h>
;;; ;; #include <arpa/inet.h>
;;; 
;;; (foreign-declare "#include \"sys/types.h\"")
;;; (foreign-declare "#include \"sys/socket.h\"")
;;; (foreign-declare "#include \"ifaddrs.h\"")
;;; (foreign-declare "#include \"arpa/inet.h\"")
;;; 
;;; ;; get IP addresses from ALL interfaces
;;; (define get-all-ips
;;;   (foreign-safe-lambda* scheme-object ()
;;;     "
;;; 
;;; // from https://stackoverflow.com/questions/17909401/linux-c-get-default-interfaces-ip-address :
;;; 
;;; 
;;;     C_word lst = C_SCHEME_END_OF_LIST, len, str, *a;
;;; //    struct ifaddrs *ifa, *i;
;;; //    struct sockaddr *sa;
;;; 
;;;     struct ifaddrs * ifAddrStruct = NULL;
;;;     struct ifaddrs * ifa = NULL;
;;;     void * tmpAddrPtr = NULL;
;;; 
;;;     if ( getifaddrs(&ifAddrStruct) != 0)
;;;       C_return(C_SCHEME_FALSE);
;;; 
;;; //    for (i = ifa; i != NULL; i = i->ifa_next) {
;;;     for (ifa = ifAddrStruct; ifa != NULL; ifa = ifa->ifa_next) {
;;;         if (ifa->ifa_addr->sa_family==AF_INET) { // Check it is
;;;             // a valid IPv4 address
;;;             tmpAddrPtr = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
;;;             char addressBuffer[INET_ADDRSTRLEN];
;;;             inet_ntop(AF_INET, tmpAddrPtr, addressBuffer, INET_ADDRSTRLEN);
;;; //            printf(\"%s IP Address %s\\n\", ifa->ifa_name, addressBuffer);
;;;             len = strlen(addressBuffer);
;;;             a = C_alloc(C_SIZEOF_PAIR + C_SIZEOF_STRING(len));
;;;             str = C_string(&a, len, addressBuffer);
;;;             lst = C_a_pair(&a, str, lst);
;;;         } 
;;; 
;;; //        else if (ifa->ifa_addr->sa_family==AF_INET6) { // Check it is
;;; //            // a valid IPv6 address
;;; //            tmpAddrPtr = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
;;; //            char addressBuffer[INET6_ADDRSTRLEN];
;;; //            inet_ntop(AF_INET6, tmpAddrPtr, addressBuffer, INET6_ADDRSTRLEN);
;;; ////            printf(\"%s IP Address %s\\n\", ifa->ifa_name, addressBuffer);
;;; //            len = strlen(addressBuffer);
;;; //            a = C_alloc(C_SIZEOF_PAIR + C_SIZEOF_STRING(len));
;;; //            str = C_string(&a, len, addressBuffer);
;;; //            lst = C_a_pair(&a, str, lst);
;;; //       }
;;; 
;;; //       else {
;;; //         printf(\" not an IPv4 address\\n\");
;;; //       }
;;; 
;;;     }
;;; 
;;;     freeifaddrs(ifa);
;;;     C_return(lst);
;;; 
;;; "))
;;; 
;;; ;; Change this to bias for addresses with a reasonable broadcast value?
;;; ;;
;;; (define (ip-pref-less? a b)
;;;   (let* ((rate (lambda (ipstr)
;;;                  (regex-case ipstr
;;;                              ( "^127\\." _ 0 )
;;;                              ( "^(10\\.0|192\\.168\\.)\\..*" _ 1 )
;;;                              ( else 2 ) ))))
;;;     (< (rate a) (rate b))))
;;;   
;;; 
;;; (define (get-my-best-address)
;;;   (let ((all-my-addresses (get-all-ips))
;;;         ;;(all-my-addresses-old (vector->list (hostinfo-addresses (hostname->hostinfo (get-host-name)))))
;;;         )
;;;     (cond
;;;      ((null? all-my-addresses)
;;;       (get-host-name))                                          ;; no interfaces?
;;;      ((eq? (length all-my-addresses) 1)
;;;       (car all-my-addresses))                      ;; only one to choose from, just go with it
;;;      
;;;      (else
;;;       (car (sort all-my-addresses ip-pref-less?)))
;;;      ;; (else 
;;;      ;;  (ip->string (car (filter (lambda (x)                      ;; take any but 127.
;;;      ;;    			 (not (eq? (u8vector-ref x 0) 127)))
;;;      ;;    		       all-my-addresses))))
;;; 
;;;      )))
;;; 
;;; (define (get-all-ips-sorted)
;;;   (sort (get-all-ips) ip-pref-less?))
;;; 
;;;