| ︙ | | |
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
|
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
|
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
+
+
+
+
+
+
+
-
+
-
+
|
(let ((varval (string-split varpair "=")))
(if (eq? (length varval) 2)
(let ((var (car varval))
(val (cadr varval)))
(debug:print 1 "Adding pre-var/val " var " = " val " to the environment")
(setenv var val)))))
varpairs)))
(for-each
(lambda (varval)
(let ((var (car varval))
(val (cadr varval)))
(if val
(setenv var val)
(begin
(debug:print 0 "ERROR: required variable " var " does not have a valid value. Exiting")
(exit)))))
(list
(setenv "MT_TEST_RUN_DIR" work-area)
(setenv "MT_TEST_NAME" test-name)
(setenv "MT_ITEM_INFO" (conc itemdat))
(setenv "MT_RUNNAME" runname)
(setenv "MT_MEGATEST" megatest)
(setenv "MT_TARGET" target)
(setenv "MT_LINKTREE" (configf:lookup *configdat* "setup" "linktree"))
(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_MEGATEST" megatest)
(list "MT_TARGET" target)
(list "MT_LINKTREE" (configf:lookup *configdat* "setup" "linktree"))))
(if mt-bindir-path (setenv "PATH" (conc (getenv "PATH") ":" mt-bindir-path)))
;; (change-directory top-path)
(if (not (setup-for-run))
(begin
(debug:print 0 "Failed to setup, exiting")
;; (sqlite3:finalize! db)
;; (sqlite3:finalize! tdb)
(exit 1)))
;; Can setup as client for server mode now
;; (client:setup)
(change-directory *toppath*)
(set-megatest-env-vars run-id) ;; these may be needed by the launching process
(set-megatest-env-vars run-id inkeys: keys inkeyvals: keyvals) ;; these may be needed by the launching process
(change-directory work-area)
(set-run-config-vars run-id keyvals target) ;; (db:get-target db run-id))
;; environment overrides are done *before* the remaining critical envars.
(alist->env-vars env-ovrd)
(set-megatest-env-vars run-id)
(set-megatest-env-vars run-id inkeys: keys inkeyvals: keyvals)
(set-item-env-vars itemdat)
(save-environment-as-files "megatest")
;; open-run-close not needed for test-set-meta-info
(tests:set-full-meta-info #f test-id run-id 0 work-area)
;; (tests:test-set-status! test-id "REMOTEHOSTSTART" "n/a" (args:get-arg "-m") #f)
(tests:test-force-state-status! test-id "REMOTEHOSTSTART" "n/a")
|
| ︙ | | |
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
|
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
|
-
-
-
-
-
+
+
+
+
+
-
-
+
+
|
((eq? rollup-status 0) 'pass)
(else 'fail)))
(next-status (cond
((eq? overall-status 'pass) this-step-status)
((eq? overall-status 'warn)
(if (eq? this-step-status 'fail) 'fail 'warn))
(else 'fail)))
(next-state "RUNNING")
;; (cond
;; ((null? tal) ;; more to run?
;; "COMPLETED")
;; (else "RUNNING"))
(next-state ;; "RUNNING") ;; WHY WAS THIS CHANGED TO NOT USE (null? tal) ??
(cond
((null? tal) ;; more to run?
"COMPLETED")
(else "RUNNING")))
)
(debug:print 4 "Exit value received: " (vector-ref exit-info 2) " logpro-used: " logpro-used
" this-step-status: " this-step-status " overall-status: " overall-status
" next-status: " next-status " rollup-status: " rollup-status)
(case next-status
((warn)
(set! rollup-status 2)
;; NB// test-set-status! does rdb calls under the hood
(tests:test-set-status! test-id next-state "WARN"
(if (eq? this-step-status 'warn) "Logpro warning found" #f)
#f))
((pass)
(tests:test-set-status! test-id next-state "PASS" #f #f))
(else ;; 'fail
(set! rollup-status 1) ;; force fail
(tests:test-set-status! test-id next-state "FAIL" (conc "Failed at step " stepname) #f)
(set! rollup-status 1) ;; force fail, this used to be next-state but that doesn't make sense. should always be "COMPLETED"
(tests:test-set-status! test-id "COMPLETED" "FAIL" (conc "Failed at step " stepname) #f)
))))
(if (and (steprun-good? logpro-used (vector-ref exit-info 2))
(not (null? tal)))
(loop (car tal) (cdr tal) stepname)))
(debug:print 4 "WARNING: a prior step failed, stopping at " ezstep))))))))
(monitorjob (lambda ()
(let* ((start-seconds (current-seconds))
|
| ︙ | | |
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
|
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
|
-
+
+
-
+
+
+
+
-
+
|
;; (car processes))
;; (system (conc "kill -9 -" pid))))
(begin
(debug:print 0 "WARNING: Request received to kill job but problem with process, attempting to kill manager process")
(tests:test-set-status! test-id "KILLED" "FAIL"
(args:get-arg "-m") #f)
(sqlite3:finalize! tdb)
(exit 1))))
(exit 1) ;; IS THIS NECESSARY OR WISE???
)))
(set! kill-tries (+ 1 kill-tries))
(mutex-unlock! m)))
;; (sqlite3:finalize! db)
(if keep-going
(begin
(thread-sleep! 3) ;; (+ 3 (random 6))) ;; add some jitter to the call home time to spread out the db accesses
(if keep-going
(loop (calc-minutes)))))))))) ;; NOTE: Checking twice for keep-going is intentional
(loop (calc-minutes)))))))
(tests:update-central-meta-info test-id (get-cpu-load) (get-df (current-directory))(calc-minutes) #f #f)))) ;; NOTE: Checking twice for keep-going is intentional
(th1 (make-thread monitorjob "monitor job"))
(th2 (make-thread runit "run job")))
(set! job-thread th2)
(thread-start! th1)
(thread-start! th2)
(thread-join! th2)
(set! keep-going #f)
(thread-join! th1)
;; (thread-sleep! 1)
;; (thread-terminate! th1) ;; Not sure if this is a good idea
(thread-sleep! 1) ;; give thread th1 a chance to be done TODO: Verify this is needed. At 0.1 I was getting fail to stop, increased to total of 1.1 sec.
;; (tests:update-central-meta-info test-id cpuload diskfree minutes #f #f)
(mutex-lock! m)
(let* ((item-path (item-list->path itemdat))
;; only state and status needed - use lazy routine
(testinfo (cdb:get-test-info-by-id *runremote* test-id))) ;; )) ;; run-id test-name item-path)))
(testinfo (cdb:remote-run db:get-testinfo-state-status #f test-id))) ;;;(cdb:get-test-info-by-id *runremote* test-id))) ;; )) ;; run-id test-name item-path)))
;; Am I completed?
(if (member (db:test-get-state testinfo) '("REMOTEHOSTSTART" "RUNNING")) ;; NOTE: It should *not* be REMOTEHOSTSTART but for reasons I don't yet understand it sometimes gets stuck in that state ;; (not (equal? (db:test-get-state testinfo) "COMPLETED"))
(let ((new-state (if kill-job? "KILLED" "COMPLETED") ;; (if (eq? (vector-ref exit-info 2) 0) ;; exited with "good" status
;; "COMPLETED"
;; (db:test-get-state testinfo))) ;; else preseve the state as set within the test
)
(new-status (cond
|
| ︙ | | |
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
|
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
|
-
+
|
" - creating link from: " test-path "\n"
" to: " lnktarget)
;; If there is already a symlink delete it and recreate it.
(handle-exceptions
exn
(begin
(debug:print 0 "ERROR: Failed to re-create link " linktarget ((condition-property-accessor 'exn 'message) exn) ", exiting")
(debug:print 0 "ERROR: Failed to re-create link " lnktarget ((condition-property-accessor 'exn 'message) exn) ", exiting")
(exit))
(if (symbolic-link? lnktarget) (delete-file lnktarget))
(if (not (file-exists? lnktarget)) (create-symbolic-link test-path lnktarget)))))
;; I suspect this section was deleting test directories under some
;; wierd sitations? This doesn't make sense - reenabling the rm -f
;; I honestly don't remember *why* this chunk was needed...
|
| ︙ | | |
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
|
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
|
-
+
+
|
(launcher (config-lookup *configdat* "jobtools" "launcher"))
(runscript (config-lookup test-conf "setup" "runscript"))
(ezsteps (> (length (hash-table-ref/default test-conf "ezsteps" '())) 0)) ;; don't send all the steps, could be big
(diskspace (config-lookup test-conf "requirements" "diskspace"))
(memory (config-lookup test-conf "requirements" "memory"))
(hosts (config-lookup *configdat* "jobtools" "workhosts"))
(remote-megatest (config-lookup *configdat* "setup" "executable"))
(run-time-limit (configf:lookup test-conf "requirements" "runtimelim"))
(run-time-limit (or (configf:lookup test-conf "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 (let* ((lm (car (argv)))
(dir (pathname-directory lm))
(exe (pathname-strip-directory lm)))
|
| ︙ | | |