Megatest

Diff
Login

Differences From Artifact [9ff8002b65]:

To Artifact [74747a8909]:


203
204
205
206
207
208
209
210
211
212
213
214
215
216


217
218
219
220
221
222
223
			   (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)

    (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







|





|
>
>







203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
			   (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
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
             (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)
          (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)
        (when do-sync
          ;;(with-output-to-file (conc (getenv "MT_TEST_RUN_DIR") "/last-loadinfo.log" #:append)
          ;;  (lambda () (pp (list (current-seconds) new-cpu-load new-disk-free (calc-minutes)))))
          (common:telemetry-log "zombie" (conc  "launch:monitor-job - dosync started at "(current-seconds)))
          (tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f)
          (common:telemetry-log "zombie" (conc "launch:monitor-job - dosync finished at "(current-seconds))))
        
	(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?







|








|






|
<
<
<
|
<
<







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
             (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?
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
	    (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))) ;; 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")







|







307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
	    (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")
463
464
465
466
467
468
469
470


471
472
473
474
475
476
477
				(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.
	     ((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


	      (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"))







|
>
>







460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
				(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"))
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
           (else #f))))
    (when do-scan?
      (debug:print 1 *default-log-port* "INFO: search and mark zombie tests")
      (rmt:set-var key (current-seconds))
      (rmt:find-and-mark-incomplete run-id #f))))













;; 1. look though disks list for disk with most space
;; 2. create run dir on disk, path name is meaningful
;; 3. create link from run dir to megatest runs area 
;; 4. remotely run the test on allocated host
;;    - could be ssh to host from hosts table (update regularly with load)
;;    - could be netbatch
;;      (launch-test db (cadr status) test-conf))
(define (launch-test test-id run-id run-info keyvals runname test-conf test-name test-path itemdat params)
  (mutex-lock! *launch-setup-mutex*) ;; setting variables and processing the testconfig is NOT thread-safe, reuse the launch-setup mutex
  (let* ( ;; (lock-key        (conc "test-" test-id))
	;; (got-lock        (let loop ((lock        (rmt:no-sync-get-lock lock-key))
	;; 			     (expire-time (+ (current-seconds) 15))) ;; give up on getting the lock and steal it after 15 seconds
	;; 		    (if (car lock)
	;; 			#t
	;; 			(if (> (current-seconds) expire-time)
	;; 			    (begin
	;; 			      (debug:print-info 0 *default-log-port* "Timed out waiting for a lock to launch test " keyvals " " runname " " test-name " " test-path)
	;; 			      (rmt:no-sync-del! lock-key) ;; destroy the lock
	;; 			      (loop (rmt:no-sync-get-lock lock-key) expire-time)) ;; 
	;; 			    (begin
	;; 			      (thread-sleep! 1)
	;; 			      (loop (rmt:no-sync-get-lock lock-key) expire-time))))))
	 (item-path       (item-list->path itemdat))
	 (contour         #f)) ;; NOT READY FOR THIS (args:get-arg "-contour")))



    (let loop ((delta        (- (current-seconds) *last-launch*))
	       (launch-delay (configf:lookup-number *configdat* "setup" "launch-delay" default: 0)))
      (if (> launch-delay delta)
	  (begin
	    (if (common:low-noise-print 1200 "test launch delay") ;; every two hours or so remind the user about launch delay.
		(debug:print-info 0 *default-log-port* "NOTE: test launches are delayed by " launch-delay " seconds. See megatest.config launch-delay setting to adjust.")) ;; launch of " test-name " for " (- launch-delay delta) " seconds"))
	    (thread-sleep! (- launch-delay delta))
	    (loop (- (current-seconds) *last-launch*) launch-delay))))
    (change-directory *toppath*)
    (alist->env-vars ;; consolidate this code with the code in megatest.scm for "-execute", *maybe* - the longer they are set the longer each launch takes (must be non-overlapping with the vars)
     (append
      (list
       (list "MT_RUN_AREA_HOME" *toppath*)
       (list "MT_TEST_NAME" test-name)
       (list "MT_RUNNAME"   runname)
       (list "MT_ITEMPATH"  item-path)
       (list "MT_CONTOUR"   contour)
       )
      itemdat))







    (let* ((tregistry       (tests:get-all)) ;; third param (below) is system-allowed
           ;; for tconfig, why do we allow fallback to test-conf?
	   (tconfig         (or (tests:get-testconfig test-name item-path tregistry #t force-create: #t)
				(begin
                                  (debug:print 0 *default-log-port* "WARNING: falling back to pre-calculated testconfig. This is likely not desired.")
                                  test-conf))) ;; force re-read now that all vars are set
	   (useshell        (let ((ush (configf:lookup *configdat* "jobtools"     "useshell")))







|
>
>
>
>
>

>
>
>










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

|
>
>
>









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







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
           (else #f))))
    (when do-scan?
      (debug:print 1 *default-log-port* "INFO: search and mark zombie tests")
      (rmt:set-var key (current-seconds))
      (rmt:find-and-mark-incomplete run-id #f))))


(defstruct launch:ajt
  (vars '())
  (exekey #f)
  (host-type #f)
  (test-sig  #f)
  (cmdline   #f))

;; append vars
(define (launch:ajt-add-vars dat vars)
  (launch:ajt-vars-set! dat (append (launch:ajt-vars dat) vars)))

;; 1. look though disks list for disk with most space
;; 2. create run dir on disk, path name is meaningful
;; 3. create link from run dir to megatest runs area 
;; 4. remotely run the test on allocated host
;;    - could be ssh to host from hosts table (update regularly with load)
;;    - could be netbatch
;;      (launch-test db (cadr status) test-conf))
(define (launch-test test-id run-id run-info keyvals runname test-conf test-name test-path itemdat params)
  (mutex-lock! *launch-setup-mutex*) ;; setting variables and processing the testconfig is NOT thread-safe, reuse the launch-setup mutex
  (let* (;; locking code removed from here commented out and pasted at end of file












	 (item-path       (item-list->path itemdat))
	 (contour         #f)                         ;; NOT READY FOR THIS (args:get-arg "-contour")))
	 ;; launcher-mode will be 'adjutant or 'normal
	 (launcher-mode   (string->symbol (or (configf:lookup *configdat* "jobtools" "mode") "normal")))
	 (ajtdat          (make-launch:ajt)))
    (let loop ((delta        (- (current-seconds) *last-launch*))
	       (launch-delay (configf:lookup-number *configdat* "setup" "launch-delay" default: 0)))
      (if (> launch-delay delta)
	  (begin
	    (if (common:low-noise-print 1200 "test launch delay") ;; every two hours or so remind the user about launch delay.
		(debug:print-info 0 *default-log-port* "NOTE: test launches are delayed by " launch-delay " seconds. See megatest.config launch-delay setting to adjust.")) ;; launch of " test-name " for " (- launch-delay delta) " seconds"))
	    (thread-sleep! (- launch-delay delta))
	    (loop (- (current-seconds) *last-launch*) launch-delay))))
    (change-directory *toppath*)

    (let ((var-list (append
		     (list
		      (list "MT_RUN_AREA_HOME" *toppath*)
		      (list "MT_TEST_NAME" test-name)
		      (list "MT_RUNNAME"   runname)
		      (list "MT_ITEMPATH"  item-path)
		      (list "MT_CONTOUR"   contour)
		      )
		     itemdat)))
       ;; consolidate this code with the code in megatest.scm for
       ;; "-execute", *maybe* - the longer they are set the longer
       ;; each launch takes (must be non-overlapping with the vars)
      (alist->env-vars var-list)
      ;; the var-list into the ajtdat adjutant record whether it is needed or not.
      (launch:ajt-add-vars ajtdat var-list))
    
    (let* ((tregistry       (tests:get-all)) ;; third param (below) is system-allowed
           ;; for tconfig, why do we allow fallback to test-conf?
	   (tconfig         (or (tests:get-testconfig test-name item-path tregistry #t force-create: #t)
				(begin
                                  (debug:print 0 *default-log-port* "WARNING: falling back to pre-calculated testconfig. This is likely not desired.")
                                  test-conf))) ;; force re-read now that all vars are set
	   (useshell        (let ((ush (configf:lookup *configdat* "jobtools"     "useshell")))
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
	   (subrun          (> (length (hash-table-ref/default tconfig "subrun"  '())) 0)) ;; send a flag to process a subrun
	   ;; (diskspace       (configf:lookup tconfig   "requirements" "diskspace"))
	   ;; (memory          (configf:lookup tconfig   "requirements" "memory"))
	   ;; (hosts           (configf:lookup *configdat* "jobtools"     "workhosts")) ;; I'm pretty sure this was never completed
	   (remote-megatest (configf:lookup *configdat* "setup" "executable"))
	   (run-time-limit  (or (configf:lookup  tconfig   "requirements" "runtimelim")
				(configf:lookup  *configdat* "setup" "runtimelim")))
	   ;; FIXME SOMEDAY: not good how this is so obtuse, this hack is to 
	   ;;                allow running from dashboard. Extract the path
	   ;;                from the called megatest and convert dashboard
	   ;;             	  or dboard to megatest
	   (local-megatest  (common:find-local-megatest))
	   #;(local-megatest  (let* ((lm  (car (argv)))
				   (dir (pathname-directory lm))
				   (exe (pathname-strip-directory lm)))

			      (conc (if dir (conc dir "/") "")
				    (case (string->symbol exe)
				      ((dboard)    "../megatest")
				      ((mtest)     "../megatest")
				      ((dashboard) "megatest")
				      (else exe)))))
	   (launcher        (common:get-launcher *configdat* test-name item-path)) ;; (configf:lookup *configdat* "jobtools"     "launcher"))


	   (test-sig        (conc (common:get-testsuite-name) ":" test-name ":" item-path)) ;; (item-list->path itemdat))) ;; test-path is the full path including the item-path
	   (work-area       #f)
	   (toptest-work-area #f) ;; for iterated tests the top test contains data relevant for all
	   (diskpath   #f)
	   (cmdparms   #f)
	   (fullcmd    #f) ;; (define a (with-output-to-string (lambda ()(write x))))
	   (mt-bindir-path #f)
	   (testinfo   (rmt:get-test-info-by-id run-id test-id))
	   (mt_target  (string-intersperse (map cadr keyvals) "/"))
	   (debug-param (append (if (args:get-arg "-debug")  (list "-debug" (args:get-arg "-debug")) '())
				(if (args:get-arg "-logging")(list "-logging") '())
				(if (configf:lookup *configdat* "misc" "profilesw")
				    (list (configf:lookup *configdat* "misc" "profilesw"))
				    '()))))









      ;; (if hosts (set! hosts (string-split hosts)))
      ;; set the megatest to be called on the remote host
      (if (not remote-megatest)(set! remote-megatest local-megatest)) ;; "megatest"))
      (set! mt-bindir-path (pathname-directory remote-megatest))
      (if launcher (set! launcher (string-split launcher)))
      ;; set up the run work area for this test
      (if (and (args:get-arg "-preclean") ;; user has requested to preclean for this run
	       (not (member (db:test-get-rundir testinfo)(list "n/a" "/tmp/badname")))) ;; n/a is a placeholder and thus not a read dir
	  (begin
	    (debug:print-info 0 *default-log-port* "attempting to preclean directory " (db:test-get-rundir testinfo) " for test " test-name "/" item-path)
	    (runs:remove-test-directory testinfo 'remove-data-only))) ;; remove data only, do not perturb the record
      







<
<
<
<

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













>
>
>
>
>
>
>
>
>




|







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
	   (subrun          (> (length (hash-table-ref/default tconfig "subrun"  '())) 0)) ;; send a flag to process a subrun
	   ;; (diskspace       (configf:lookup tconfig   "requirements" "diskspace"))
	   ;; (memory          (configf:lookup tconfig   "requirements" "memory"))
	   ;; (hosts           (configf:lookup *configdat* "jobtools"     "workhosts")) ;; I'm pretty sure this was never completed
	   (remote-megatest (configf:lookup *configdat* "setup" "executable"))
	   (run-time-limit  (or (configf:lookup  tconfig   "requirements" "runtimelim")
				(configf:lookup  *configdat* "setup" "runtimelim")))




	   (local-megatest  (common:find-local-megatest))



	   (launcher        (let ((l (common:get-launcher *configdat* test-name item-path launcher-mode)))
			      (if (string? l)
				  (string-split l)





				  l))) ;; some nonhomogenuity here. '(cmd param1 param2 ...) OR '(host-type launcher)
	    ;; (item-list->path itemdat))) ;; test-path is the full path including the item-path
	   (test-sig        (conc (common:get-testsuite-name) ":" test-name ":" item-path))
	   (work-area       #f)
	   (toptest-work-area #f) ;; for iterated tests the top test contains data relevant for all
	   (diskpath   #f)
	   (cmdparms   #f)
	   (fullcmd    #f) ;; (define a (with-output-to-string (lambda ()(write x))))
	   (mt-bindir-path #f)
	   (testinfo   (rmt:get-test-info-by-id run-id test-id))
	   (mt_target  (string-intersperse (map cadr keyvals) "/"))
	   (debug-param (append (if (args:get-arg "-debug")  (list "-debug" (args:get-arg "-debug")) '())
				(if (args:get-arg "-logging")(list "-logging") '())
				(if (configf:lookup *configdat* "misc" "profilesw")
				    (list (configf:lookup *configdat* "misc" "profilesw"))
				    '()))))
      ;; save the test-sig in the ajtdat record
      (launch:ajt-test-sig-set! ajtdat test-sig)
      ;; go ahead and figure out if we have a host-type from the
      ;; launcher call above and save it in the ajtdat record
      (if (and (eq? launcher-mode 'adjutant)
	       (list? launcher)
	       (> (length launcher) 1))
	  (launch:ajt-host-type-set! ajtdat (car launcher)))
 
      ;; (if hosts (set! hosts (string-split hosts)))
      ;; set the megatest to be called on the remote host
      (if (not remote-megatest)(set! remote-megatest local-megatest)) ;; "megatest"))
      (set! mt-bindir-path (pathname-directory remote-megatest))
      ;; (if launcher (set! launcher (string-split launcher)))           ;; yuk!
      ;; set up the run work area for this test
      (if (and (args:get-arg "-preclean") ;; user has requested to preclean for this run
	       (not (member (db:test-get-rundir testinfo)(list "n/a" "/tmp/badname")))) ;; n/a is a placeholder and thus not a read dir
	  (begin
	    (debug:print-info 0 *default-log-port* "attempting to preclean directory " (db:test-get-rundir testinfo) " for test " test-name "/" item-path)
	    (runs:remove-test-directory testinfo 'remove-data-only))) ;; remove data only, do not perturb the record
      
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
					(list 'target    mt_target)
					(list 'contour   contour)
					(list 'runtlim   (if run-time-limit (common:hms-string->seconds run-time-limit) #f))
					(list 'env-ovrd  (hash-table-ref/default *configdat* "env-override" '())) 
					(list 'set-vars  (if params (hash-table-ref/default params "-setvars" #f)))
					(list 'runname   runname)
					(list 'mt-bindir-path mt-bindir-path))))))))


      
      ;; clean out step records from previous run if they exist
      ;; (rmt:delete-test-step-records run-id test-id)
      ;; if the dir does not exist we may have a itempath where individual variables are a path, launch anyway
      (if (common:file-exists? work-area)
	  (change-directory work-area)) ;; so that log files from the launch process don't clutter the test dir
      (cond

       ;; ((and launcher hosts) ;; must be using ssh hostname
       ;;    (set! fullcmd (append launcher (car hosts)(list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param)))
       ;; (set! fullcmd (append launcher (car hosts)(list remote-megatest test-sig "-execute" cmdparms))))

       (launcher
	(set! fullcmd (append launcher (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param)))
       ;; (set! fullcmd (append launcher (list remote-megatest test-sig "-execute" cmdparms))))
       (else
	(if (not useshell)(debug:print 0 *default-log-port* "WARNING: internal launching will not work well without \"useshell yes\" in your [jobtools] section"))
	(set! fullcmd (append (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param (list (if useshell "&" ""))))))
      ;; (set! fullcmd (list remote-megatest test-sig "-execute" cmdparms (if useshell "&" "")))))
      (if (args:get-arg "-xterm")(set! fullcmd (append fullcmd (list "-xterm"))))
      (debug:print 1 *default-log-port* "Launching " work-area)
      ;; set pre-launch-env-vars before launching, keep the vars in prevvals and put the envionment back when done
      (debug:print 4 *default-log-port* "fullcmd: " fullcmd)
      (set! *last-launch* (current-seconds)) ;; all that junk above takes time, set this as late as possible.
      (let* ((commonprevvals (alist->env-vars
			      (hash-table-ref/default *configdat* "env-override" '())))
	     (miscprevvals   (alist->env-vars ;; consolidate this code with the code in megatest.scm for "-execute"
			      (append (list (list "MT_TEST_RUN_DIR" work-area)
					    (list "MT_TEST_NAME" test-name)
					    (list "MT_ITEM_INFO" (conc itemdat)) 
					    (list "MT_RUNNAME"   runname)
					    (list "MT_TARGET"    mt_target)
					    (list "MT_ITEMPATH"  item-path)
					    )
				      itemdat)))


	     (testprevvals   (alist->env-vars
			      (hash-table-ref/default tconfig "pre-launch-env-overrides" '())))
	     ;; Launchwait defaults to true, must override it to turn off wait
	     (launchwait     (if (equal? (configf:lookup *configdat* "setup" "launchwait") "no") #f #t))
	     (launch-results-prev (apply (if launchwait ;; BB: TODO: refactor this to examine return code of launcher, if nonzero, set state to launch failed.



					     process:cmd-run-with-stderr-and-exitcode->list
					     process-run)
					 (if useshell
					     (let ((cmdstr (string-intersperse fullcmd " ")))
					       (if launchwait
						   cmdstr
						   (conc cmdstr " >> mt_launch.log 2>&1 &")))
					     (car fullcmd))
					 (if useshell
					     '()
					     (cdr fullcmd))))
             (success        (if launchwait (equal? 0 (cadr launch-results-prev)) #t))
             (launch-results (if launchwait (car launch-results-prev) launch-results-prev)))
























        (if (not success)
            (tests:test-set-status! run-id test-id "COMPLETED" "DEAD" "launcher failed; exited non-zero; check mt_launch.log" #f)) ;; (if launch-results launch-results "FAILED"))
        (mutex-unlock! *launch-setup-mutex*) ;; yes, really should mutex all the way to here. Need to put this entire process into a fork.
	;; (rmt:no-sync-del! lock-key)         ;; release the lock for starting this test
	(if (not launchwait) ;; give the OS a little time to allow the process to start
	    (thread-sleep! 0.01))
	(with-output-to-file "mt_launch.log"
	  (lambda ()
	    (print "LAUNCHCMD: " (string-intersperse fullcmd " "))
	    (if (list? launch-results)







>
>






|
>
|
|
<
>


<



|





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


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


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

<







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
					(list 'target    mt_target)
					(list 'contour   contour)
					(list 'runtlim   (if run-time-limit (common:hms-string->seconds run-time-limit) #f))
					(list 'env-ovrd  (hash-table-ref/default *configdat* "env-override" '())) 
					(list 'set-vars  (if params (hash-table-ref/default params "-setvars" #f)))
					(list 'runname   runname)
					(list 'mt-bindir-path mt-bindir-path))))))))
      ;; save the cmdparms in the ajtdat
      (launch:ajt-exekey-set! ajtdat cmdparms)
      
      ;; clean out step records from previous run if they exist
      ;; (rmt:delete-test-step-records run-id test-id)
      ;; if the dir does not exist we may have a itempath where individual variables are a path, launch anyway
      (if (common:file-exists? work-area)
	  (change-directory work-area)) ;; so that log files from the launch process don't clutter the test dir

      ;; save the command line for adjutant mode (might never be needed but best to assemble it here)
      (launch:ajt-cmdline-set! ajtdat (string-intersperse
				       (append (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param)))

      (cond       
       (launcher
	(set! fullcmd (append launcher (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param)))

       (else
	(if (not useshell)(debug:print 0 *default-log-port* "WARNING: internal launching will not work well without \"useshell yes\" in your [jobtools] section"))
	(set! fullcmd (append (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param (list (if useshell "&" ""))))))
      
      (if (args:get-arg "-xterm")(set! fullcmd (append fullcmd (list "-xterm"))))
      (debug:print 1 *default-log-port* "Launching " work-area)
      ;; set pre-launch-env-vars before launching, keep the vars in prevvals and put the envionment back when done
      (debug:print 4 *default-log-port* "fullcmd: " fullcmd)
      (set! *last-launch* (current-seconds)) ;; all that junk above takes time, set this as late as possible.

      (let* ((env-override-vars  (hash-table-ref/default *configdat* "env-override" '()))
	     (commonprevvals     (alist->env-vars env-override-vars))
	     (misc-vars          (append (list (list "MT_TEST_RUN_DIR" work-area)
					       (list "MT_TEST_NAME" test-name)
					       (list "MT_ITEM_INFO" (conc itemdat)) 
					       (list "MT_RUNNAME"   runname)
					       (list "MT_TARGET"    mt_target)
					       (list "MT_ITEMPATH"  item-path))

					 itemdat))
	     (miscprevvals   (alist->env-vars misc-vars));; consolidate this code with the code in megatest.scm for "-execute"
	     (test-vars      (hash-table-ref/default tconfig "pre-launch-env-overrides" '()))
	     (testprevvals   (alist->env-vars test-vars))
			      
	     ;; Launchwait defaults to true, must override it to turn off wait
	     (launchwait     (if (equal? (configf:lookup *configdat* "setup" "launchwait") "no") #f #t))
	     ;; BB: TODO: refactor this to examine return code of launcher, if nonzero, set state to launch failed.
	     (launch-results-prev (if (eq? launcher-mode 'adjutant)
				      '(#t 0) ;; just some fake data to fool downstream but non-applicable code
				      (apply (if launchwait
						 process:cmd-run-with-stderr-and-exitcode->list
						 process-run)
					     (if useshell
						 (let ((cmdstr (string-intersperse fullcmd " ")))
						   (if launchwait
						       cmdstr
						       (conc cmdstr " >> mt_launch.log 2>&1 &")))
						 (car fullcmd))
					     (if useshell
						 '()
						 (cdr fullcmd)))))
             (success        (if launchwait (equal? 0 (cadr launch-results-prev)) #t))
             (launch-results (if launchwait (car launch-results-prev) launch-results-prev)))

	(launch:ajt-add-vars ajtdat env-override-vars)
	(launch:ajt-add-vars ajtdat misc-vars)
	(launch:ajt-add-vars ajtdat test-vars)

	;; if in adjutant mode we register the job in the jobs_queue
	;; then fire off an adjutant runner
	;;
	(if (eq? launcher-mode 'adjutant)
	    (let* ((adjutant-runner-cmd (append (cdr launcher)
						(list remote-megatest "-adjutant"
						      (launch:ajt-host-type ajtdat)
						      "-start-dir" *toppath*)))
		   (adj-cmd     (conc (string-intersperse (map conc adjutant-runner-cmd) " ")
				      "&")))         
	      (rmt:no-sync-add-job
	       (launch:ajt-host-type  ajtdat)
	       (conc (launch:ajt-vars ajtdat))
	       (launch:ajt-exekey     ajtdat)
	       (launch:ajt-cmdline    ajtdat))
	      (print "adj-cmd: " adj-cmd)
	      (system adj-cmd)
	      ))
	
	(if (not success)
            (tests:test-set-status! run-id test-id "COMPLETED" "DEAD" "launcher failed; exited non-zero; check mt_launch.log" #f)) ;; (if launch-results launch-results "FAILED"))

	;; (rmt:no-sync-del! lock-key)         ;; release the lock for starting this test
	(if (not launchwait) ;; give the OS a little time to allow the process to start
	    (thread-sleep! 0.01))
	(with-output-to-file "mt_launch.log"
	  (lambda ()
	    (print "LAUNCHCMD: " (string-intersperse fullcmd " "))
	    (if (list? launch-results)
1589
1590
1591
1592
1593
1594
1595




1596
1597
1598
1599
1600
1601
1602
	      ;; but this hack will work! Thanks go to Alan Post of the Chicken email list
	      ;; NB// Is this still needed? Should be safe to go back to "exit" now?
	      (process-signal (current-process-id) signal/kill)
	      ))
	(alist->env-vars miscprevvals)
	(alist->env-vars testprevvals)
	(alist->env-vars commonprevvals)




	launch-results))
    (change-directory *toppath*)
    (thread-sleep! (configf:lookup-number *configdat* "setup" "inter-test-delay" default: 0.0))))

;; recover a test where the top controlling mtest may have died
;;
(define (launch:recover-test run-id test-id)







>
>
>
>







1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
	      ;; but this hack will work! Thanks go to Alan Post of the Chicken email list
	      ;; NB// Is this still needed? Should be safe to go back to "exit" now?
	      (process-signal (current-process-id) signal/kill)
	      ))
	(alist->env-vars miscprevvals)
	(alist->env-vars testprevvals)
	(alist->env-vars commonprevvals)
	;; yes, really should mutex all the way to here. Need to put this entire process into a fork.
	;; the unlock previously was further up. This seemed wrong as we should not proceed until the
	;; vars have been reset.
	(mutex-unlock! *launch-setup-mutex*)
	launch-results))
    (change-directory *toppath*)
    (thread-sleep! (configf:lookup-number *configdat* "setup" "inter-test-delay" default: 0.0))))

;; recover a test where the top controlling mtest may have died
;;
(define (launch:recover-test run-id test-id)
1618
1619
1620
1621
1622
1623
1624
















		     (read-symbolic-link (conc "/proc/" pid "/cwd"))
		     #f)))
    ;; now wait on that process if all is correct
    ;; periodically update the db with runtime
    ;; when the process exits look at the db, if still RUNNING after 10 seconds set
    ;; state/status appropriately
    (process-wait pid)))























>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
		     (read-symbolic-link (conc "/proc/" pid "/cwd"))
		     #f)))
    ;; now wait on that process if all is correct
    ;; periodically update the db with runtime
    ;; when the process exits look at the db, if still RUNNING after 10 seconds set
    ;; state/status appropriately
    (process-wait pid)))


 ;; (lock-key        (conc "test-" test-id))
	;; (got-lock        (let loop ((lock        (rmt:no-sync-get-lock lock-key))
	;; 			     (expire-time (+ (current-seconds) 15))) ;; give up on getting the lock and steal it after 15 seconds
	;; 		    (if (car lock)
	;; 			#t
	;; 			(if (> (current-seconds) expire-time)
	;; 			    (begin
	;; 			      (debug:print-info 0 *default-log-port* "Timed out waiting for a lock to launch test " keyvals " " runname " " test-name " " test-path)
	;; 			      (rmt:no-sync-del! lock-key) ;; destroy the lock
	;; 			      (loop (rmt:no-sync-get-lock lock-key) expire-time)) ;; 
	;; 			    (begin
	;; 			      (thread-sleep! 1)
	;; 			      (loop (rmt:no-sync-get-lock lock-key) expire-time))))))