Megatest

# 
#     You should have received a copy of the GNU General Public License
#     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

TODO
====

WW38
. Add test_rundat to no-sync ==> correction, put in <testdir>/.meta/test-run.dat
. Add STATE/STATUS transitions to .meta/test-run.dat or similar
. Swizzle update-test-rundat to operate on no-sync
. Swizzle update-run-duration, -uname-host and cpuload-diskfree to no-sync
. On state/status change update tests table with duration

WW15
. fill newview matrix with data, filter pipeline gui elements
. improve [script], especially indent handling

WW16
. split db into megatest.db (runs etc.) db/<something>.db
. release basic newview implementation

WW18
. release split db implementation
. mtutil calls from dashboard (for remote control)
. logs browser (esp. for surfacing mtutil related activities)

WW19
. break command line into sections; all, run control, queries, utilities etc.
. pull in ftfplan (not integrated, just code pulled in)

WW20
. ./configure => ubuntu, sles11, sles12, rh7

. Add output flushing in teamcity support
. Switch to using simple runs query everywhere
. Add end_time to runs and add a rollup call that sets state, status and end_time

Future
. Switch to scsh-process pipeline management for job execution/control
. Use call-with-environment-variables more.

				   (lambda (a b)
				     (< (hash-table-ref all-files a)(hash-table-ref all-files b))))
			     (- num-logs max-allowed))))
	    (for-each
	     (lambda (file)
	       (let* ((fullname (conc "logs/" file)))
		 (if (directory? fullname)
		     (debug:print-info 0 *default-log-port* fullname " in logs directory is a directory! Cannot rotate it, it is best to not put subdirectories in the logs dir.")
		     (handle-exceptions
		      exn
		      (debug:print-error 0 *default-log-port* "failed to remove " fullname ", exn=" exn)
		      (delete-file* fullname)))))
	     files)
	    (debug:print-info 0 *default-log-port* "Deleted " (length files) " files from logs, keeping " max-allowed " files."))))))
  

;;
(define (dboard:common-set-tabdat! commondat tabnum tabdat)
  (hash-table-set!
   (dboard:commondat-tabdats commondat)
   tabnum
   tabdat))

(define *updater-running* #f) ;; move this into one of the stucts

;; gets and calls updater list based on curr-tab-num
;;
(define (dboard:common-run-curr-updaters commondat #!key (tab-num #f))
  (if (dboard:common-get-tabdat commondat tab-num: tab-num) ;; only update if there is a tabdat
      (let* ((tnum     (or tab-num (dboard:commondat-curr-tab-num commondat)))
	     (updaters (hash-table-ref/default (dboard:commondat-updaters commondat)
					       tnum

	 (curr-updaters (hash-table-ref/default (dboard:commondat-updaters commondat) tnum '())))
    (hash-table-set! (dboard:commondat-updaters commondat)
		     tnum
		     (cons updater curr-updaters))))

;; data for each specific tab goes here
;;
(defstruct dboard:tabdat
  ;; flags
  ((already-running #f)                  : boolean)     ;; the updater is already running. skip
  ;; runs
  ((allruns         '())                 : list)        ;; list of dboard:rundat records
  ((allruns-by-id    (make-hash-table))  : hash-table)  ;; hash of run-id -> dboard:rundat records
  ((done-runs       '())                 : list)        ;; list of runs already drawn
  ((not-done-runs   '())                 : list)        ;; list of runs not yet drawn
  (header            #f)                                ;; header for decoding the run records
  (keys              #f)                                ;; keys for this run (i.e. target components)

;;
;;    NOTE: Yes, this is used
;;
(define (dboard:get-tests-for-run-duplicate tabdat run-id run testnamepatt key-vals)
  (let* ((start-time   (current-seconds))
	 (access-mode  (dboard:tabdat-access-mode tabdat))
         (num-to-get   (string->number (or (configf:lookup *configdat* "setup" "num-tests-to-get")
                                           "60"))) ;; was 200, which is fine in a normal run area.
	 (states       (hash-table-keys (dboard:tabdat-state-ignore-hash tabdat)))
	 (statuses     (hash-table-keys (dboard:tabdat-status-ignore-hash tabdat)))
         (do-not-use-db-file-timestamps #t) ;; (configf:lookup *configdat* "setup" "do-not-use-db-file-timestamps")) ;; this still hosts runs-summary-tab
         (do-not-use-query-timestamps   #t) ;; (configf:lookup *configdat* "setup" "do-not-use-query-timestamps")) ;; this no longer troubles runs-summary-tab
	 (sort-info    (get-curr-sort))
	 (sort-by      (vector-ref sort-info 1))
	 (sort-order   (vector-ref sort-info 2))

    ;;
    ;; (debug:print 0 *default-log-port* "got-all: " got-all " multi-get: " multi-get " num-to-get: " num-to-get " (length tmptests): " (length tmptests) " db-modified: " db-modified " db-mod-time: " db-mod-time " db-path: " db-path)
    (if got-all
	(begin
	  (dboard:rundat-last-update-set!     run-dat (- start-time 2))
	  (dboard:rundat-run-data-offset-set! run-dat 0))
	(begin
	  ;;; (thread-sleep! 0.25) ;; give the rest of the gui some time to update. <-- this did NOT help
	  (dboard:rundat-run-data-offset-set! run-dat
					      (+ num-to-get (dboard:rundat-run-data-offset run-dat)))))

    (for-each 
     (lambda (tdat)
       (let ((test-id (db:test-get-id tdat))
	     (state   (db:test-get-state tdat)))

;; this calls dboard:get-tests-for-run-duplicate for each run
;;
;; create a virtual table of all the tests
;; keypatts: ( (KEY1 "abc%def")(KEY2 "%") )
;;
(define (dboard:update-rundat tabdat runnamepatt numruns testnamepatt keypatts)
  (dboard:tabdat-already-running-set! tabdat #t)
  (let* (;; (already-running  (dboard:tabdat-already-running tabdat))
	 (access-mode      (dboard:tabdat-access-mode tabdat))
         (keys             (dboard:tabdat-keys tabdat)) ;; (db:dispatch-query access-mode rmt:get-keys db:get-keys)))
	 (last-runs-update (- (dboard:tabdat-last-runs-update tabdat) 2))
         (allruns          (rmt:get-runs runnamepatt numruns (dboard:tabdat-start-run-offset tabdat) keypatts))
         ;;(allruns-tree (rmt:get-runs-by-patt (dboard:tabdat-keys tabdat) "%" #f #f #f #f))
         (allruns-tree    (rmt:get-runs-by-patt keys "%" #f #f #f #f 0)) ;; last-runs-update));;'("id" "runname")
	 (header      (db:get-header allruns))
	 (runs        (db:get-rows   allruns)) ;; RA => Filtered as per runpatt selected

		   (elapsed-time (- (current-seconds) start-time)))
	      (if (null? all-test-ids)
		  (hash-table-delete! (dboard:tabdat-allruns-by-id tabdat) run-id)
		  (hash-table-set!    (dboard:tabdat-allruns-by-id tabdat) run-id run-struct))
	      (if (or (null? tal)
		      (> elapsed-time 2)) ;; stop loading data after 5 seconds, on the next call more data *should* be loaded since get-tests-for-run uses last update
		  (begin
		    #;(when (> elapsed-time 2)   
                      (debug:print 0 *default-log-port* "NOTE: updates are taking a long time, " elapsed-time "s elapsed.")
                      (let* ((old-val (iup:attribute *tim* "TIME"))
                             (new-val (number->string (inexact->exact (floor (* 2  (string->number old-val)))))))
                        (if (< (string->number new-val) 5000)
                            (begin
			      (debug:print 0 *default-log-port* "NOTE: increasing poll interval from "old-val" to "new-val)
			      (iup:attribute-set! *tim* "TIME" new-val)))))



		    (dboard:tabdat-allruns-set! tabdat new-res)
		    maxtests)
		  (if (> (dboard:rundat-run-data-offset run-struct) 0)
		      (loop run tal new-res newmaxtests) ;; not done getting data for this run
		      (loop (car tal)(cdr tal) new-res newmaxtests)))))))
    (dboard:tabdat-filters-changed-set! tabdat #f)
    (dboard:update-tree tabdat runs-hash header tb)
    (dboard:tabdat-already-running-set! tabdat #f)))

(define *collapsed* (make-hash-table))

(define (toggle-hide lnum uidat) ; fulltestname)
  (let* ((btn (vector-ref (dboard:uidat-get-lftcol uidat) lnum))
	 (fulltestname (iup:attribute btn "TITLE"))
	 (parts        (string-split fulltestname "("))

            sort-lb))) 
	)

        ;; insert extra widget here
        (if extra-widget
            extra-widget
            (iup:hbox)) ;; empty widget




        )))

     (let* ((status-toggles (map (lambda (status)
				   (iup:toggle (conc status)
					       #:fontsize 8 ;; btn-fontsz ;; "10"
					       ;; #:expand "HORIZONTAL"
					       #:action   (lambda (obj val)

;;  removing the tabdat-values proc 
;;
;; (define (tabdat-values tabdat)

;; runs update-rundat using the various filters from the gui
;;
(define (dashboard:do-update-rundat tabdat)
  ;; this seems like a good place to check for already running and skip if so
  ;;
  ;; (set! *updater-running* #t)
;;(if (dboard:tabdat-already-running tabdat)
;;    (begin
;;	(debug:print-info 0 *default-log-port* "Dashboard overloaded - updates will be slow, skipping update.")
;;	(dboard:tabdat-target tabdat))
  (dboard:update-rundat
   tabdat
   (hash-table-ref/default (dboard:tabdat-searchpatts tabdat) "runname" "%")
   (dboard:tabdat-numruns tabdat)
   (hash-table-ref/default (dboard:tabdat-searchpatts tabdat) "test-name" "%/%")
   ;; generate key patterns from the target stored in tabdat
   (let* ((dbkeys (dboard:tabdat-dbkeys tabdat)))
     (let ((fres   (if (dboard:tabdat-target tabdat)
		       (let ((ptparts (append (dboard:tabdat-target tabdat)(make-list (length dbkeys) "%"))))
			 (map (lambda (k v)(list k v)) dbkeys ptparts))
		       (let ((res '()))
			 (for-each (lambda (key)
				     (if (not (equal? key "runname"))
					 (let ((val (hash-table-ref/default (dboard:tabdat-searchpatts tabdat) key #f)))
					   (if val (set! res (cons (list key val) res))))))
				   dbkeys)
			 res))))
       fres)))
  #;(set! *updater-running* #f))

(define (dashboard:runs-tab-updater commondat tab-num)
  (debug:catch-and-dump 
   (lambda ()
     (let* ((tabdat (dboard:common-get-tabdat commondat tab-num: tab-num))
	    (dbkeys (dboard:tabdat-dbkeys tabdat)))
       (dashboard:do-update-rundat tabdat)

	 commondat 
	 (lambda ()
	   (dashboard:runs-tab-updater commondat 1))
	 tab-num: 2)
	(iup:callback-set! *tim*
			   "ACTION_CB"
			   (lambda (time-obj)
			     (if (not *updater-running*)
				 (begin
				   ;; (mutex-lock! (dboard:commondat-update-mutex commondat))
				   ;; (set! update-is-running (dboard:commondat-updating commondat))
				   ;;(if (not update-is-running)
				   ;;	   (dboard:commondat-updating-set! commondat #t))
				   ;;(mutex-unlock! (dboard:commondat-update-mutex commondat))
				   ;;(if (not update-is-running) ;; we know that the update was not running and we now have a lock on doing an update
				   ;;	   (begin
				   (set! *updater-running* #t)
				   (dboard:common-run-curr-updaters commondat) ;; (dashboard:run-update commondat)
				   (set! *updater-running* #f)
				   ;; (mutex-lock! (dboard:commondat-update-mutex commondat))
				   ;;  (dboard:commondat-updating-set! commondat #f)
				   ;;   (mutex-unlock! (dboard:commondat-update-mutex commondat)))
				   ))
			     1))))
      
      (let ((th1 (make-thread (lambda ()
				(thread-sleep! 1)
				(dboard:common-run-curr-updaters commondat 0) ;; force update of summary tab 
				) "update buttons once"))

                              id           INTEGER PRIMARY KEY,
                              test_id      INTEGER,
                              state        TEXT DEFAULT 'new',
                              status       TEXT DEFAULT 'n/a',
                              archive_type TEXT DEFAULT 'bup',
                              du           INTEGER,
                              archive_path TEXT);")))
         ;; (print "creating trigges from init") 
        (db:create-triggers db)    
     db)) ;; )

;;======================================================================
;; A R C H I V E S
;;======================================================================

  (let* ((run-ids (db:get-all-run-ids mtdb)))
    (for-each 
     (lambda (run-id)
       (let ((testrecs (db:get-all-tests-info-by-run-id mtdb run-id)))
	 (db:prep-megatest.db-adj-test-ids (db:dbdat-get-db mtdb) run-id testrecs)))
     run-ids)))

;; Get test data using test_id, run-id is not used - but it will be!
;; 
(define (db:get-test-info-by-id dbstruct run-id test-id)
  (db:with-db
   dbstruct
   #f ;; run-id
   #f
   (lambda (db)

<td class="tableblock halign-left valign-top" ><p class="tableblock monospaced"></p></td>
<td class="tableblock halign-left valign-top" ><p class="tableblock monospaced">( key1 key2 &#8230; )</p></td>
<td class="tableblock halign-left valign-top" ><p class="tableblock monospaced"></p></td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="sect1">
<h2 id="_test_plan">Test Plan</h2>
<div class="sectionbody">
<div class="sect2">
<h3 id="_tests">Tests</h3>
<div class="paragraph"><p>itemwait|33</p></div>
<div class="paragraph"><p>rerun-downstream-item|20</p></div>
<div class="paragraph"><p>rerunclean|20</p></div>
<div class="paragraph"><p>fullrun|18</p></div>
<div class="paragraph"><p>goodtests|18</p></div>
<div class="paragraph"><p>kill-rerun|17</p></div>
<div class="paragraph"><p>items-runconfigvars|16</p></div>
<div class="paragraph"><p>ro_test|16</p></div>
<div class="paragraph"><p>runconfig-tests|16</p></div>
<div class="paragraph"><p>env-pollution|13</p></div>
<div class="paragraph"><p>itemmap|11</p></div>
<div class="paragraph"><p>testpatt_envvar|10</p></div>
<div class="paragraph"><p>toprun|10</p></div>
<div class="paragraph"><p>chained-waiton|8</p></div>
<div class="paragraph"><p>skip-on-fileexists|8</p></div>
<div class="paragraph"><p>killrun_preqfail|7</p></div>
<div class="paragraph"><p>subrun|6</p></div>
<div class="paragraph"><p>dependencies|5</p></div>
<div class="paragraph"><p>itemwait-simple|4</p></div>
<div class="paragraph"><p>rollup|4</p></div>
<div class="paragraph"><p>end-of-run|3</p></div>
<div class="paragraph"><p>killrun|3</p></div>
<div class="paragraph"><p>listener|3</p></div>
<div class="paragraph"><p>test2|3</p></div>
<div class="paragraph"><p>testpatt|3</p></div>
<div class="paragraph"><p>env-pollution-usecacheno|2</p></div>
<div class="paragraph"><p>set-values|2
envvars|1
listruns-tests|1
subrun-usecases|1</p></div>
</div>
</div>
</div>
<div class="sect1">
<h2 id="_megatest_internals">Megatest Internals</h2>
<div class="sectionbody">
<div class="imageblock graphviz">
<div class="content">
<img src="server.png" alt="server.png">

</div>
</div>
</div>
<div id="footnotes"><hr></div>
<div id="footer">
<div id="footer-text">
Version 1.5<br>
Last updated 2020-09-08 08:39:29 PDT
 2020-08-22 12:47:36 MST
</div>
</div>
</body>
</html>

// :leveloffset: 0

include::writing_tests.txt[]

include::howto.txt[]

include::reference.txt[]

include::testplan.txt[]

Megatest Internals
------------------

["graphviz", "server.png"]
----------------------------------------------------------------------
include::server.dot[]

// This file is part of Megatest.
// 
//     Megatest is free software: you can redistribute it and/or modify
//     it under the terms of the GNU General Public License as published by
//     the Free Software Foundation, either version 3 of the License, or
//     (at your option) any later version.
// 
//     Megatest is distributed in the hope that it will be useful,
//     but WITHOUT ANY WARRANTY; without even the implied warranty of
//     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
//     GNU General Public License for more details.
// 
//     You should have received a copy of the GNU General Public License
//     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

// Copyright 2006-2020, Matthew Welland.

Test Plan
---------

Tests
~~~~~

itemwait|33

rerun-downstream-item|20

rerunclean|20

fullrun|18

goodtests|18

kill-rerun|17

items-runconfigvars|16

ro_test|16

runconfig-tests|16

env-pollution|13

itemmap|11

testpatt_envvar|10

toprun|10

chained-waiton|8

skip-on-fileexists|8

killrun_preqfail|7

subrun|6

dependencies|5

itemwait-simple|4

rollup|4

end-of-run|3

killrun|3

listener|3

test2|3

testpatt|3

env-pollution-usecacheno|2

set-values|2
envvars|1
listruns-tests|1
subrun-usecases|1

			   (round 
			    (- 
			     (current-seconds) 
			     start-seconds)))))
	 (kill-tries 0))
    ;; (tests:set-full-meta-info #f test-id run-id (calc-minutes) work-area)
    ;; (tests:set-full-meta-info test-id run-id (calc-minutes) work-area)
    (tests:set-full-meta-info #f test-id run-id (calc-minutes) work-area 10 update-db: #t)

    (let loop ((minutes   (calc-minutes))
	       (cpu-load  (alist-ref 'adj-core-load (common:get-normalized-cpu-load #f)))
	       (disk-free (get-df (current-directory)))
               (last-sync (current-seconds)))
      ;; (common:telemetry-log "zombie" (conc "launch:monitor-job -
      ;; top of loop encountered at "(current-seconds)" with
      ;; last-sync="last-sync))
      (let* ((over-time     (> (current-seconds) (+ last-sync update-period)))
             (new-cpu-load  (let* ((load  (alist-ref 'adj-core-load (common:get-normalized-cpu-load #f)))
                                   (delta (abs (- load cpu-load))))
                              (if (> delta 0.1) ;; don't bother updating with small changes
                                  load
                                  #f)))
             (new-disk-free (let* ((df    (if over-time ;; only get df every 30 seconds

             (do-sync       (or new-cpu-load new-disk-free over-time))

             (test-info   (rmt:get-test-info-by-id run-id test-id))
             (state       (db:test-get-state test-info))
             (status      (db:test-get-status test-info))
             (kill-reason  "no kill reason specified")
             (kill-job?    #f))
        #;(common:telemetry-log "zombie" (conc "launch:monitor-job - decision time encountered at "(current-seconds)" with last-sync="last-sync" do-sync="do-sync" over-time="over-time" update-period="update-period))
        (cond
         ((test-get-kill-request run-id test-id)
          (set! kill-reason "KILLING TEST since received kill request (KILLREQ)")
          (set! kill-job? #t))
         ((and runtlim (> (- (current-seconds) start-seconds) runtlim))
          (set! kill-reason (conc "KILLING TEST DUE TO TIME LIMIT EXCEEDED! Runtime=" (- (current-seconds) start-seconds) " seconds, limit=" runtlim))
          (set! kill-job? #t))
         ((equal? status "DEAD")
          (tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f update-db: #t)
          (rmt:set-state-status-and-roll-up-items run-id test-id 'foo "RUNNING" "n/a" "was marked dead; really still running.")
          ;;(set! kill-reason "KILLING TEST because it was marked as DEAD by launch:handle-zombie-tests (might indicate really overloaded server or else overzealous setup.deadtime)") ;; MARK RUNNING
          (set! kill-job? #f)))

        (debug:print 4 *default-log-port* "cpu: " new-cpu-load " disk: " new-disk-free " last-sync: " last-sync " do-sync: " do-sync)
        (launch:handle-zombie-tests run-id)
        (if do-sync ;; save meta data about the running of this test



	    (tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f))


	(if kill-job? 
	    (begin
              (debug:print-info 0 *default-log-port* "proceeding to kill test: "kill-reason)
	      (mutex-lock! m)
	      ;; NOTE: The pid can change as different steps are run. Do we need handshaking between this
	      ;;       section and the runit section? Or add a loop that tries three times with a 1/4 second
	      ;;       between tries?

	    (begin
	      (thread-sleep! 3) ;; (+ 3 (random 6))) ;; add some jitter to the call home time to spread out the db accesses
	      (if (hash-table-ref/default misc-flags 'keep-going #f)  ;; keep originals for cpu-load and disk-free unless they change more than the allowed delta
		  (loop (calc-minutes)
                        (or new-cpu-load cpu-load)
                        (or new-disk-free disk-free)
                        (if do-sync (current-seconds) last-sync)))))))
    (tests:update-central-meta-info run-id test-id (get-cpu-load) (get-df (current-directory))(calc-minutes) #f #f update-db: #t))) ;; NOTE: Checking twice for keep-going is intentional


(define (launch:execute encoded-cmd)
  (let* ((cmdinfo    (common:read-encoded-string encoded-cmd))
	 (tconfigreg #f))
    (setenv "MT_CMDINFO" encoded-cmd)
    ;;(bb-check-path msg: "launch:execute incoming")

				(db:test-get-host        test-info)
				(begin
				  (debug:print 0 *default-log-port* "ERROR: failed to find a record for test-id " test-id ", exiting.")
				  (exit))))
		 (test-pid  (db:test-get-process_id  test-info)))
	    (cond
             ;; -mrw- I'm removing KILLREQ from this list so that a test in KILLREQ state is treated as a "do not run" flag.
	     ((or (member (db:test-get-state test-info) '("INCOMPLETE" "KILLED" "UNKNOWN" "STUCK")) ;; prior run of this test didn't complete, go ahead and try to rerun
		  (and (equal? (db:test-get-state test-info) "COMPLETED")                           ;; completed/abort => rerun if asked
		       (member (db:test-get-status test-info) '("ABORT"))))
	      (debug:print 0 *default-log-port* "INFO: test is INCOMPLETE or KILLED, treat this execute call as a rerun request")
	      ;; (tests:test-force-state-status! run-id test-id "REMOTEHOSTSTART" "n/a")

              (rmt:general-call 'set-test-start-time #f test-id)
              (rmt:test-set-state-status run-id test-id "REMOTEHOSTSTART" "n/a" #f)
	      ) ;; prime it for running
	     ((member (db:test-get-state test-info) '("RUNNING" "REMOTEHOSTSTART"))

;;     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

;; Always use two or four digit decimal
;; 1.01, 1.02...1.10,1.11,1.1101 ... 1.99,2.00..

;; (declare (unit megatest-version))

(define megatest-version 1.6569)

(declare (uses tdb))
(declare (uses mt))
(declare (uses api))
(declare (uses tasks)) ;; only used for debugging.
(declare (uses env))
(declare (uses diff-report))

(declare (uses mutils))
(import mutils)

(declare (uses adjutant))
(import adjutant)

;; (declare (uses ftail))
;; (import ftail)

     readline apropos json http-client directory-utils typed-records matchable
     http-client srfi-18 extras format)

;; Added for csv stuff - will be removed
;;
(use sparse-vectors)

;; (require-library mutils)

(define *usage-log-file* #f)    ;; put path to file for logging usage in this var in the ~/.megatestrc file
(define *usage-use-seconds* #t) ;; for Epoc seconds in usage logging change this to #t in ~/.megatestrc file

;; load the ~/.megatestrc file, put (use trace)(trace-call-sites #t)(trace function-you-want-to-trace) in this file
;;
(let ((debugcontrolf (conc (get-environment-variable "HOME") "/.megatestrc")))

(use format typed-records) ;; RADT => purpose of json format??

(declare (unit rmt))
(declare (uses api))
(declare (uses http-transport))
(include "common_records.scm")
(include "db_records.scm")

;; (declare (uses rmtmod))

;; (import rmtmod)

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

	      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-query-last-call-time* 0)
(define *rmt-query-last-rest-time* 0) ;; last time there was at least a 1/2 second rest - giving other processes access to the db

;; NOTE: This query rest algorythm will not adapt to long query times. REDESIGN NEEDED. TODO. FIXME.
;;
(define (rmt:query-rest)
  (let* ((now (current-milliseconds)))
    (cond
     ((> (- now *rmt-query-last-call-time*) 500)  ;; it's been a while since last query - no need to rest
      (set! *rmt-query-last-rest-time*  now)
      (set! *rmt-query-last-call-time*  now))
     ((> (- now *rmt-query-last-rest-time*) 5000) ;; no natural rests have happened
      (debug:print 0 *default-log-port* "query rest needed. blocking for 1/2 second.")
      (thread-sleep! 0.5) ;; force a rest of a half second
      (set! *rmt-query-last-rest-time* now)
      (set! *rmt-query-last-call-time* now))
     (else ;; sufficient rests have occurred, just record the last query time
      (set! *rmt-query-last-call-time* now)))))

;; RA => e.g. usage (rmt:send-receive 'get-var #f (list varname))
;;
(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)))
  (if (not (equal? (configf:lookup *configdat* "setup" "query-rest") "no"))
      (rmt:query-rest))
  
  (if (> attemptnum 2)
      (debug:print 0 *default-log-port* "INFO: attemptnum in rmt:send-receive is " attemptnum))
    
  (cond
   ((> attemptnum 2) (thread-sleep! 0.05))
   ((> attemptnum 10) (thread-sleep! 0.5))
   ((> attemptnum 20) (thread-sleep! 1)))

;; Just some syntatic sugar
(define (rmt:register-test run-id test-name item-path)
  (rmt:general-call 'register-test run-id run-id test-name item-path))

(define (rmt:get-test-id run-id testname item-path)
  (rmt:send-receive 'get-test-id run-id (list run-id testname item-path)))

;; run-id is NOT used - but it will be! 
;;
(define (rmt:get-test-info-by-id run-id test-id)
  (if (number? test-id)
      (let* ((testdat  (rmt:send-receive 'get-test-info-by-id run-id (list run-id test-id)))
	     (trundatf (conc (db:test-get-rundir testdat) "/.mt_data/test-run.dat")))
	;; now we can update a couple fields from the filesystem
	(if (and (db:test-get-rundir testdat)
		 (file-exists? trundatf))
	    (let* ((duration   (db:test-get-run_duration testdat))
		   (event-time (db:test-get-event_time   testdat))
		   (last-touch (file-modification-time trundatf)))
	      (db:test-set-run_duration! testdat (max duration (- last-touch event-time)))))
	testdat)
      (begin
	(debug:print 0 *default-log-port* "WARNING: Bad data handed to rmt:get-test-info-by-id run-id=" run-id ", test-id=" test-id)
	(print-call-chain (current-error-port))
	#f)))

(define (rmt:test-get-rundir-from-test-id run-id test-id)
  (rmt:send-receive 'test-get-rundir-from-test-id run-id (list run-id test-id)))

                                 (debug:print 0 *default-log-port* "INFO: syncer thread NOT sleeping ; maybe time-to-exit...")))

                               (if (not (configf:lookup *configdat* "server" "disable-db-snapshot"))
                                   (common:snapshot-file mtdbfile subdir: ".db-snapshot"))
		               (delete-file* staging-file)
		               (let* ((start-time (current-milliseconds))
                                      (res (system sync-cmd))
                                      (dbbackupfile (conc mtdbfile ".backup"))
                                      (res2 
                                       (cond
                                        ((eq? 0 res )
                                         (if (file-exists? dbbackupfile)
		                           (delete-file* dbbackupfile)
                                         )
                                         (if (eq? 0 (file-size sync-log))
                                             (delete-file sync-log))
		                         (system (conc "/bin/mv " staging-file " " mtdbfile))
                                         
                                         (set! last-sync-seconds (/ (- (current-milliseconds) start-time) 1000))
                                         (set! off-time (calculate-off-time
                                                         last-sync-seconds

	 (lambda (count)
	   (set! res count))
	 tdb
	 "SELECT count(id) FROM test_rundat;")
	res))
  0)

;; 
(define (tests:update-central-meta-info run-id test-id cpuload diskfree minutes uname hostname #!key (update-db #f)(tmpfree #f))
  (if (get-environment-variable "MT_TEST_RUN_DIR")
      (let* ((dest-dir (conc (get-environment-variable "MT_TEST_RUN_DIR") "/.mt_data"))
	     (or-dash  (lambda (instr)
			 (cond
			  ((not instr) "") ;; #f -> blank, indicates value unchanged since last measurement taken
			  ((string? instr)(if (string-search " " instr) (conc "\"" instr "\"") instr))
			  (else instr))))
	     (file-new (not (directory-exists? dest-dir))))
	(if file-new (create-directory dest-dir #t))
	(let* ((outp (open-output-file (conc dest-dir "/test-run.dat") #:append)))
	  (with-output-to-port outp
	    (lambda ()
	      (if file-new
		  (print "epoch_time,run_id,test_id,cpuload,diskfree,tmpfree,run_minutes,hostname,uname"))
	      (print (current-seconds) "," (or-dash run-id)   "," (or-dash test-id)  ","
		     (or-dash cpuload) "," (or-dash diskfree) "," (or-dash tmpfree)  ","
		     (or-dash minutes) "," (or-dash hostname) ","
		     (or-dash uname)))) ;; put uname last as it has spaces in it
	  (close-output-port outp)))
      (begin
	(rmt:general-call 'update-test-rundat run-id test-id (current-seconds) (or cpuload -1)(or diskfree -1) -1 (or minutes -1))))
  (if update-db
      (begin
	(if (and cpuload diskfree)
	    (rmt:general-call 'update-cpuload-diskfree run-id cpuload diskfree test-id))
	(if minutes 
	    (rmt:general-call 'update-run-duration run-id minutes test-id))
	(if (and uname hostname)
	    (rmt:general-call 'update-uname-host run-id uname hostname test-id)))))
  
;; This one is for running with no db access (i.e. via rmt: internally)
(define (tests:set-full-meta-info db test-id run-id minutes work-area remtries #!key (update-db #f))
;; (define (tests:set-full-meta-info test-id run-id minutes work-area)
;;  (let ((remtries 10))
  (let* ((cpuload  (get-cpu-load))
	 (diskfree (get-df (current-directory)))
	 (tmpfree  (get-df "/tmp"))
	 (uname    (get-uname "-srvpio"))
	 (hostname (get-host-name)))
    (tests:update-central-meta-info run-id test-id cpuload diskfree minutes uname hostname update-db: update-db tmpfree: tmpfree)))
    























	 
;;======================================================================
;; A R C H I V I N G
;;======================================================================

(define (test:archive db test-id)
  #f)