| ︙ | | | ︙ | |
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
|
(status (conc (configf:lookup dat "final" "exit-status") ": " (if msg msg "no message")))
(else #f)))
#f)))
(define (launch:runstep ezstep run-id test-id exit-info m tal testconfig)
(let* ((stepname (car ezstep)) ;; do stuff to run the step
(stepinfo (cadr ezstep))
(stepparts (string-match (regexp "^(\\{([^\\}]*)\\}\\s*|)(.*)$") stepinfo))
(stepparms (list-ref stepparts 2)) ;; for future use, {VAR=1,2,3}, run step for each
(stepcmd (list-ref stepparts 3))
(script "") ; "#!/bin/bash\n") ;; yep, we depend on bin/bash FIXME!!!\
(logpro-file (conc stepname ".logpro"))
(html-file (conc stepname ".html"))
(dat-file (conc stepname ".dat"))
(tconfig-logpro (configf:lookup testconfig "logpro" stepname))
(logpro-used (common:file-exists? logpro-file)))
(if (and tconfig-logpro
(not logpro-used)) ;; no logpro file found but have a defn in the testconfig
(begin
(with-output-to-file logpro-file
(lambda ()
(print ";; logpro file extracted from testconfig\n"
";;")
(print tconfig-logpro)))
(set! logpro-used #t)))
;; NB// can safely assume we are in test-area directory
(debug:print 4 *default-log-port* "ezsteps:\n stepname: " stepname " stepinfo: " stepinfo " stepparts: " stepparts
" stepparms: " stepparms " stepcmd: " stepcmd)
;; ;; first source the previous environment
;; (let ((prev-env (conc ".ezsteps/" prevstep (if (string-search (regexp "csh")
;; (get-environment-variable "SHELL")) ".csh" ".sh"))))
;; (if (and prevstep (common:file-exists? prev-env))
;; (set! script (conc script "source " prev-env))))
;; call the command using mt_ezstep
;; (set! script (conc "mt_ezstep " stepname " " (if prevstep prevstep "x") " " stepcmd))
(debug:print 4 *default-log-port* "script: " script)
(rmt:teststep-set-status! run-id test-id stepname "start" "-" #f #f)
;; now launch the actual process
(call-with-environment-variables
(list (cons "PATH" (conc (get-environment-variable "PATH") ":.")))
(lambda () ;; (process-run "/bin/bash" "-c" "exec ls -l /tmp/foobar > /tmp/delme-more.log 2>&1")
(let* ((cmd (conc stepcmd " > " stepname ".log 2>&1")) ;; >outfile 2>&1
(pid (process-run "/bin/bash" (list "-c" cmd))))
(with-output-to-file "Makefile.ezsteps"
(lambda ()
(print stepname ".log :")
(print "\t" cmd)
(if (common:file-exists? (conc stepname ".logpro"))
(print "\tlogpro " stepname ".logpro " stepname ".html < " stepname ".log"))
(print)
|
>
>
>
>
|
|
>
>
>
>
>
>
>
|
>
>
|
>
>
>
>
>
|
|
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
|
(status (conc (configf:lookup dat "final" "exit-status") ": " (if msg msg "no message")))
(else #f)))
#f)))
(define (launch:runstep ezstep run-id test-id exit-info m tal testconfig)
(let* ((stepname (car ezstep)) ;; do stuff to run the step
(stepinfo (cadr ezstep))
;; (let ((info (cadr ezstep)))
;; (if (proc? info) "" info)))
;; (stepproc (let ((info (cadr ezstep)))
;; (if (proc? info) info #f)))
(stepparts (string-match (regexp "^(\\{([^\\}\\{]*)\\}\\s*|)(.*)$") stepinfo))
(stepparams (list-ref stepparts 2)) ;; for future use, {VAR=1,2,3}, run step for each
(paramparts (if (string? stepparams)
(map (lambda (x)(string-split x "=")) (string-split-fields "[^;]*=[^;]*" stepparams))
'()))
(subrun (alist-ref "subrun" paramparts equal?))
(stepcmd (list-ref stepparts 3))
(script "") ; "#!/bin/bash\n") ;; yep, we depend on bin/bash FIXME!!!\
(logpro-file (conc stepname ".logpro"))
(html-file (conc stepname ".html"))
(dat-file (conc stepname ".dat"))
(tconfig-logpro (configf:lookup testconfig "logpro" stepname))
(logpro-used (common:file-exists? logpro-file)))
(debug:print 0 *default-log-port* "stepparts: " stepparts ", stepparams: " stepparams
", paramparts: " paramparts ", subrun: " subrun ", stepcmd: " stepcmd)
(if (and tconfig-logpro
(not logpro-used)) ;; no logpro file found but have a defn in the testconfig
(begin
(with-output-to-file logpro-file
(lambda ()
(print ";; logpro file extracted from testconfig\n"
";;")
(print tconfig-logpro)))
(set! logpro-used #t)))
;; NB// can safely assume we are in test-area directory
(debug:print 4 *default-log-port* "ezsteps:\n stepname: " stepname " stepinfo: " stepinfo " stepparts: " stepparts
" stepparams: " stepparams " stepcmd: " stepcmd)
;; ;; first source the previous environment
;; (let ((prev-env (conc ".ezsteps/" prevstep (if (string-search (regexp "csh")
;; (get-environment-variable "SHELL")) ".csh" ".sh"))))
;; (if (and prevstep (common:file-exists? prev-env))
;; (set! script (conc script "source " prev-env))))
;; call the command using mt_ezstep
;; (set! script (conc "mt_ezstep " stepname " " (if prevstep prevstep "x") " " stepcmd))
(debug:print 4 *default-log-port* "script: " script)
(rmt:teststep-set-status! run-id test-id stepname "start" "-" #f #f)
;; now launch the actual process
(call-with-environment-variables
(list (cons "PATH" (conc (get-environment-variable "PATH") ":.")))
(lambda () ;; (process-run "/bin/bash" "-c" "exec ls -l /tmp/foobar > /tmp/delme-more.log 2>&1")
(let* ((cmd (conc stepcmd " > " stepname ".log 2>&1")) ;; >outfile 2>&1
(pid #f))
(let ((proc (lambda ()
(set! pid (process-run "/bin/bash" (list "-c" cmd))))))
(if subrun
(begin
(debug:print-info 0 *default-log-port* "Running without MT_.* environment variables.")
(common:without-vars proc "^MT_.*"))
(proc)))
(with-output-to-file "Makefile.ezsteps"
(lambda ()
(print stepname ".log :")
(print "\t" cmd)
(if (common:file-exists? (conc stepname ".logpro"))
(print "\tlogpro " stepname ".logpro " stepname ".html < " stepname ".log"))
(print)
|
| ︙ | | | ︙ | |
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
|
(tests:test-set-status! run-id test-id next-state "PASS" #f #f))
(else ;; 'fail
(launch:einf-rollup-status-set! exit-info 1) ;; (vector-set! exit-info 3 1) ;; force fail, this used to be next-state but that doesn't make sense. should always be "COMPLETED"
(tests:test-set-status! run-id test-id "COMPLETED" "FAIL" (conc "Failed at step " stepname) #f)
)))
logpro-used))
(define (launch:manage-steps run-id test-id item-path fullrunscript ezsteps test-name tconfigreg exit-info m)
;; (let-values
;; (((pid exit-status exit-code)
;; (run-n-wait fullrunscript)))
;; (tests:test-set-status! test-id "RUNNING" "n/a" #f #f)
;; Since we should have a clean slate at this time there is no need to do
;; any of the other stuff that tests:test-set-status! does. Let's just
;; force RUNNING/n/a
|
|
|
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
|
(tests:test-set-status! run-id test-id next-state "PASS" #f #f))
(else ;; 'fail
(launch:einf-rollup-status-set! exit-info 1) ;; (vector-set! exit-info 3 1) ;; force fail, this used to be next-state but that doesn't make sense. should always be "COMPLETED"
(tests:test-set-status! run-id test-id "COMPLETED" "FAIL" (conc "Failed at step " stepname) #f)
)))
logpro-used))
(define (launch:manage-steps run-id test-id item-path fullrunscript ezsteps subrun test-name tconfigreg exit-info m)
;; (let-values
;; (((pid exit-status exit-code)
;; (run-n-wait fullrunscript)))
;; (tests:test-set-status! test-id "RUNNING" "n/a" #f #f)
;; Since we should have a clean slate at this time there is no need to do
;; any of the other stuff that tests:test-set-status! does. Let's just
;; force RUNNING/n/a
|
| ︙ | | | ︙ | |
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
|
(if (eq? pid-val 0)
(begin
(thread-sleep! 2)
(loop (+ i 1)))
)))))
;; then, if runscript ran ok (or did not get called)
;; do all the ezsteps (if any)
(if ezsteps
(let* ((testconfig ;; (read-config (conc work-area "/testconfig") #f #t environ-patt: "pre-launch-env-vars")) ;; FIXME??? is allow-system ok here?
;; NOTE: it is tempting to turn off force-create of testconfig but dynamic
;; ezstep names need a full re-eval here.
(tests:get-testconfig test-name item-path tconfigreg #t force-create: #t)) ;; 'return-procs)))
(ezstepslst (if (hash-table? testconfig)
(hash-table-ref/default testconfig "ezsteps" '())
#f)))
(if testconfig
(hash-table-set! *testconfigs* test-name testconfig) ;; cached for lazy reads later ...
(begin
(launch:setup)
(debug:print 0 *default-log-port* "WARNING: no testconfig found for " test-name " in search path:\n "
(string-intersperse (tests:get-tests-search-path *configdat*) "\n "))))
;; after all that, still no testconfig? Time to abort
(if (not testconfig)
(begin
(debug:print-error 0 *default-log-port* "Failed to resolve megatest.config, runconfigs.config and testconfig issues. Giving up now")
(exit 1)))
(if (not (common:file-exists? ".ezsteps"))(create-directory ".ezsteps"))
;; if ezsteps was defined then we are sure to have at least one step but check anyway
(if (not (> (length ezstepslst) 0))
(debug:print-error 0 *default-log-port* "ezsteps defined but ezstepslst is zero length")
(let loop ((ezstep (car ezstepslst))
(tal (cdr ezstepslst))
(prevstep #f))
;; check exit-info (vector-ref exit-info 1)
(if (launch:einf-exit-status exit-info) ;; (vector-ref exit-info 1)
(let ((logpro-used (launch:runstep ezstep run-id test-id exit-info m tal testconfig))
(stepname (car ezstep)))
;; if logpro-used read in the stepname.dat file
(if (and logpro-used (common:file-exists? (conc stepname ".dat")))
(launch:load-logpro-dat run-id test-id stepname))
(if (steprun-good? logpro-used (launch:einf-exit-code exit-info))
(if (not (null? tal))
(loop (car tal) (cdr tal) stepname))
(debug:print 4 *default-log-port* "WARNING: step " (car ezstep) " failed. Stopping")))
(debug:print 4 *default-log-port* "WARNING: a prior step failed, stopping at " ezstep)))))))
(define (launch:monitor-job run-id test-id item-path fullrunscript ezsteps test-name tconfigreg exit-info m work-area runtlim misc-flags)
(let* ((update-period (string->number (or (configf:lookup *configdat* "setup" "test-stats-update-period") "30")))
(start-seconds (current-seconds))
(calc-minutes (lambda ()
(inexact->exact
(round
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
|
|
|
|
|
|
>
|
|
|
|
|
|
|
|
|
|
|
|
|
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
|
(if (eq? pid-val 0)
(begin
(thread-sleep! 2)
(loop (+ i 1)))
)))))
;; then, if runscript ran ok (or did not get called)
;; do all the ezsteps (if any)
(if (or ezsteps subrun)
(let* ((testconfig ;; (read-config (conc work-area "/testconfig") #f #t environ-patt: "pre-launch-env-vars")) ;; FIXME??? is allow-system ok here?
;; NOTE: it is tempting to turn off force-create of testconfig but dynamic
;; ezstep names need a full re-eval here.
(tests:get-testconfig test-name item-path tconfigreg #t force-create: #t)) ;; 'return-procs)))
(ezstepslst (if (hash-table? testconfig)
(hash-table-ref/default testconfig "ezsteps" '())
#f)))
(if testconfig
(hash-table-set! *testconfigs* test-name testconfig) ;; cached for lazy reads later ...
(begin
(launch:setup)
(debug:print 0 *default-log-port* "WARNING: no testconfig found for " test-name " in search path:\n "
(string-intersperse (tests:get-tests-search-path *configdat*) "\n "))))
;; after all that, still no testconfig? Time to abort
(if (not testconfig)
(begin
(debug:print-error 0 *default-log-port* "Failed to resolve megatest.config, runconfigs.config and testconfig issues. Giving up now")
(exit 1)))
;; create a proc for the subrun if requested, save that proc in the ezsteps table as the last entry
;; 1. get section [runarun]
;; 2. unset MT_* vars
;; 3. fix target
;; 4. fix runname
;; 5. fix testpatt or calculate it from contour
;; 6. launch the run
;; 7. roll up the run result and or roll up the logpro processed result
(if (configf:lookup testconfig "subrun" "runwait") ;; we use runwait as the flag that a subrun is requested
(let* ((runarea (configf:lookup testconfig "subrun" "runarea"))
(passfail (configf:lookup testconfig "subrun" "passfail"))
(target (configf:lookup testconfig "subrun" "target"))
(runname (configf:lookup testconfig "subrun" "runname"))
(contour (configf:lookup testconfig "subrun" "contour"))
(testpatt (configf:lookup testconfig "subrun" "testpatt"))
(mode-patt (configf:lookup testconfig "subrun" "mode-patt"))
(tag-expr (configf:lookup testconfig "subrun" "tag-expr"))
(run-wait (configf:lookup testconfig "subrun" "runwait"))
(logpro (configf:lookup testconfig "subrun" "logpro"))
(compact-stem (string-substitute "[/*]" "_" (conc target "-" runname "-" (or testpatt mode-patt tag-expr))))
(log-file (conc compact-stem ".log"))
(mt-cmd (conc "megatest -run -target " target
" -runname " (or runname (get-environment-variable "MT_RUNNAME"))
(conc " -start-dir " (if runarea runarea *toppath*))
(if testpatt (conc " -testpatt " testpatt) "")
(if mode-patt (conc " -modepatt " mode-patt) "")
(if tag-expr (conc " -tag-expr" tag-expr) "")
(if (equal? run-wait "yes") " -run-wait " "")
" -log " log-file)))
;; change directory to runarea, create it if needed, we do NOT create the directory
;; (if runarea
;; (if (directory-exists? runarea)
;; (change-directory runarea)
;; (begin
;; (debug:print 0 *default-log-port* "ERROR: for sub-megatest run the runarea \"" runarea "\" does not exist! EXITING.")
;; (exit 1))))
;; (let ((subrun (conc *toppath* "/subrun") #t))
;; (create-directory subrun)
;; (change-directory subrun)))
;; by this point we are in the right place to run the subrun and we have a Megatest command to run
;; (filter (lambda (x)(string-match "MT_.*" (car x))) (get-environment-variables))
;; (common:without-vars mt-cmd "^MT_.*")
(debug:print-info 0 *default-log-port* "Subrun command is \"" mt-cmd "\"")
(set! ezsteps #t) ;; set the needed flag
(set! ezstepslst (append (or ezstepslst '())
(list (list "subrun" (conc "{subrun=true} " mt-cmd)))))
(configf:set-section-var testconfig "logpro" "subrun" logpro) ;; append the logpro rules to the logpro section as stepname subrun
(if runarea (configf:set-section-var testconfig "setup" "submegatest" runarea))
(configf:write-alist testconfig "testconfig.subrun")
))
;; process the ezsteps
(if ezsteps
(begin
(if (not (common:file-exists? ".ezsteps"))(create-directory ".ezsteps"))
;; if ezsteps was defined then we are sure to have at least one step but check anyway
(if (not (> (length ezstepslst) 0))
(debug:print-error 0 *default-log-port* "ezsteps defined but ezstepslst is zero length")
(let loop ((ezstep (car ezstepslst))
(tal (cdr ezstepslst))
(prevstep #f))
(debug:print-info 0 *default-log-port* "Processing ezstep \"" (string-intersperse ezstep " ") "\"")
;; check exit-info (vector-ref exit-info 1)
(if (launch:einf-exit-status exit-info) ;; (vector-ref exit-info 1)
(let ((logpro-used (launch:runstep ezstep run-id test-id exit-info m tal testconfig))
(stepname (car ezstep)))
;; if logpro-used read in the stepname.dat file
(if (and logpro-used (common:file-exists? (conc stepname ".dat")))
(launch:load-logpro-dat run-id test-id stepname))
(if (steprun-good? logpro-used (launch:einf-exit-code exit-info))
(if (not (null? tal))
(loop (car tal) (cdr tal) stepname))
(debug:print 0 *default-log-port* "WARNING: step " (car ezstep) " failed. Stopping")))
(debug:print 0 *default-log-port* "WARNING: a prior step failed, stopping at " ezstep)))))))))
(define (launch:monitor-job run-id test-id item-path fullrunscript ezsteps test-name tconfigreg exit-info m work-area runtlim misc-flags)
(let* ((update-period (string->number (or (configf:lookup *configdat* "setup" "test-stats-update-period") "30")))
(start-seconds (current-seconds))
(calc-minutes (lambda ()
(inexact->exact
(round
|
| ︙ | | | ︙ | |
421
422
423
424
425
426
427
428
429
430
431
432
433
434
|
;; (test-name sqlitespeed) (runscript runscript.rb) (db-host localhost) (run-id 1))
(let* ((testpath (assoc/default 'testpath cmdinfo)) ;; testpath is the test spec area
(top-path (assoc/default 'toppath cmdinfo))
(work-area (assoc/default 'work-area cmdinfo)) ;; work-area is the test run area
(test-name (assoc/default 'test-name cmdinfo))
(runscript (assoc/default 'runscript cmdinfo))
(ezsteps (assoc/default 'ezsteps cmdinfo))
;; (runremote (assoc/default 'runremote cmdinfo))
;; (transport (assoc/default 'transport cmdinfo)) ;; not used
;; (serverinf (assoc/default 'serverinf cmdinfo))
;; (port (assoc/default 'port cmdinfo))
(serverurl (assoc/default 'serverurl cmdinfo))
(homehost (assoc/default 'homehost cmdinfo))
(run-id (assoc/default 'run-id cmdinfo))
|
>
|
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
|
;; (test-name sqlitespeed) (runscript runscript.rb) (db-host localhost) (run-id 1))
(let* ((testpath (assoc/default 'testpath cmdinfo)) ;; testpath is the test spec area
(top-path (assoc/default 'toppath cmdinfo))
(work-area (assoc/default 'work-area cmdinfo)) ;; work-area is the test run area
(test-name (assoc/default 'test-name cmdinfo))
(runscript (assoc/default 'runscript cmdinfo))
(ezsteps (assoc/default 'ezsteps cmdinfo))
(subrun (assoc/default 'subrun cmdinfo))
;; (runremote (assoc/default 'runremote cmdinfo))
;; (transport (assoc/default 'transport cmdinfo)) ;; not used
;; (serverinf (assoc/default 'serverinf cmdinfo))
;; (port (assoc/default 'port cmdinfo))
(serverurl (assoc/default 'serverurl cmdinfo))
(homehost (assoc/default 'homehost cmdinfo))
(run-id (assoc/default 'run-id cmdinfo))
|
| ︙ | | | ︙ | |
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
|
(mt-bindir-path (assoc/default 'mt-bindir-path cmdinfo))
(keys #f)
(keyvals #f)
(fullrunscript (if (not runscript)
#f
(if (substring-index "/" runscript)
runscript ;; use unadultered if contains slashes
(let ((fulln (conc testpath "/" runscript)))
(if (and (common:file-exists? fulln)
(file-execute-access? fulln))
fulln
runscript))))) ;; assume it is on the path
) ;; (rollup-status 0)
;; On NFS it can be slow and unreliable to get needed startup information.
;; i. Check if we are on the homehost, if so, proceed
;; ii. Check if host and port passed in via CMDINFO are valid and if
;; possible use them.
(let ((bestadrs (server:get-best-guess-address (get-host-name)))
(needcare #f))
(if (equal? homehost bestadrs) ;; we are likely on the homehost
(debug:print-info 0 *default-log-port* "test " test-name " appears to be running on the homehost " homehost)
(let ((host-port (if serverurl (string-split serverurl ":") #f)))
(if (not *runremote*)(set! *runremote* (make-remote))) ;; init *runremote*
(if (string? homehost)
(if (and host-port
(> (length host-port) 1))
(let* ((host (car host-port))
(port (cadr host-port))
(start-res (http-transport:client-connect host port))
(ping-res (rmt:login-no-auto-client-setup start-res)))
(if (and start-res
ping-res)
;; (begin ;; let ((url (http-transport:server-dat-make-url start-res)))
(begin
(remote-conndat-set! *runremote* start-res)
;; (remote-server-url-set! *runremote* url)
;; (if (server:ping url)
(debug:print-info 0 *default-log-port* "connected to " host ":" port " using CMDINFO data."))
(begin
(debug:print-info 0 *default-log-port* "have CMDINFO data but failed to connect to " host ":" port)
(set! *runremote* #f))
;; (remote-conndat-set! *runremote* #f))
))
(begin
(set! *runremote* #f)
(debug:print-info 0 *default-log-port* (if host-port
(conc "received invalid host-port information " host-port)
"no host-port information received"))
;; potential for bad situation if simultaneous starting of hundreds of jobs on servers, set needcare.
(set! needcare #t)))
(begin
(set! *runremote* #f)
(debug:print-info 0 *default-log-port* "received no homehost information. Please report this to support as it should not happen.")
(set! needcare #t)))))
(if needcare ;; due to very slow NFS we will do a brute force mkdir to ensure that the directory inode it truly available on this host
(let ((logdir (conc top-path "/logs"))) ;; we'll try to create this directory
(handle-exceptions
exn
(debug:print 0 *default-log-port* "Failed to create directory " logdir " expect problems, message: " ((condition-property-accessor 'exn 'message) exn))
(create-directory logdir #t)))))
;; NFS might not have propagated the directory meta data to the run host - give it time if needed
(let loop ((count 0))
(if (or (common:file-exists? top-path)
(> count 10))
(change-directory top-path)
(begin
(debug:print 0 *default-log-port* "INFO: Not starting job yet - directory " top-path " not found")
(thread-sleep! 10)
(loop (+ count 1)))))
(launch:setup) ;; should be properly in the top-path now
(set! tconfigreg (tests:get-all))
|
|
>
>
>
>
>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
|
|
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
|
(mt-bindir-path (assoc/default 'mt-bindir-path cmdinfo))
(keys #f)
(keyvals #f)
(fullrunscript (if (not runscript)
#f
(if (substring-index "/" runscript)
runscript ;; use unadultered if contains slashes
(let ((fulln (conc work-area "/" runscript)))
(if (and (common:file-exists? fulln)
(file-execute-access? fulln))
fulln
runscript))))) ;; assume it is on the path
(check-work-area (lambda ()
;; NFS might not have propagated the directory meta data to the run host - give it time if needed
(let loop ((count 0))
(if (or (common:directory-exists? work-area)
(> count 10))
(change-directory work-area)
(begin
(debug:print 0 *default-log-port* "INFO: Not starting job yet - directory " work-area " not found")
(thread-sleep! 10)
(loop (+ count 1)))))
(if (not (string=? (common:real-path work-area)(common:real-path (current-directory))))
(begin
(debug:print 0 *default-log-port*
"INFO: we are expecting to be in directory " work-area "\n"
" but we are actually in the directory " (current-directory) "\n"
" doing another change dir.")
(change-directory work-area)))
;; spot check that the files in testpath are available. Too often NFS delays cause problems here.
(let ((files (glob (conc testpath "/*")))
(bad-files '()))
(for-each
(lambda (fullname)
(let* ((fname (pathname-strip-directory fullname))
(targn (conc work-area "/" fname)))
(if (not (file-exists? targn))
(set! bad-files (cons fname bad-files)))))
files)
(if (not (null? bad-files))
(begin
(debug:print 0 *default-log-port* "INFO: test data from " testpath " not copied properly or filesystem problems causing data to not be found. Re-running the copy command.")
(debug:print 0 *default-log-port* "INFO: missing files from " work-area ": " (string-intersperse bad-files ", "))
(launch:test-copy testpath work-area))))
;; one more time, change to the work-area directory
(change-directory work-area)))
) ;; let*
(if contour (setenv "MT_CONTOUR" contour))
;; immediated set some key variables from CMDINFO data, yes, these will be set again below ...
;;
(setenv "MT_TESTSUITENAME" areaname)
(setenv "MT_RUN_AREA_HOME" top-path)
(set! *toppath* top-path)
(change-directory *toppath*) ;; temporarily switch to the run area home
(setenv "MT_TEST_RUN_DIR" work-area)
(launch:setup) ;; should be properly in the run area home now
(if contour (setenv "MT_CONTOUR" contour))
;; immediated set some key variables from CMDINFO data, yes, these will be set again below ...
;;
(setenv "MT_TESTSUITENAME" areaname)
(setenv "MT_RUN_AREA_HOME" top-path)
(set! *toppath* top-path)
(change-directory *toppath*) ;; temporarily switch to the run area home
(setenv "MT_TEST_RUN_DIR" work-area)
(launch:setup) ;; should be properly in the run area home now
(set! tconfigreg (tests:get-all)) ;; mapping of testname => test source path
(let ((sighand (lambda (signum)
;; (signal-mask! signum) ;; to mask or not? seems to cause issues in exiting
(if (eq? signum signal/stop)
(debug:print-error 0 *default-log-port* "attempt to STOP process. Exiting."))
(set! *time-to-exit* #t)
(print "Received signal " signum ", cleaning up before exit. Please wait...")
(let ((th1 (make-thread (lambda ()
|
| ︙ | | | ︙ | |
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
|
(test-host (if test-info
(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
((member (db:test-get-state test-info) '("INCOMPLETE" "KILLED" "UNKNOWN" "KILLREQ" "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: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"))
(if (process:alive-on-host? test-host test-pid)
(debug:print-error 0 *default-log-port* "test state is " (db:test-get-state test-info) " and process " test-pid " is still running on host " test-host ", cannot proceed")
|
>
|
|
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
|
(test-host (if test-info
(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: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"))
(if (process:alive-on-host? test-host test-pid)
(debug:print-error 0 *default-log-port* "test state is " (db:test-get-state test-info) " and process " test-pid " is still running on host " test-host ", cannot proceed")
|
| ︙ | | | ︙ | |
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
|
;; one of these is defunct/redundant ...
(if (not (launch:setup force-reread: #t))
(begin
(debug:print 0 *default-log-port* "Failed to setup, exiting")
;; (sqlite3:finalize! db)
;; (sqlite3:finalize! tdb)
(exit 1)))
(change-directory *toppath*)
;; NOTE: Current order is to process runconfigs *before* setting the MT_ vars. This
;; seems non-ideal but could well break stuff
;; BUG? BUG? BUG?
(let ((rconfig (full-runconfigs-read)) ;; (read-config (conc *toppath* "/runconfigs.config") #f #t sections: (list "default" target))))
(wconfig (read-config "waivers.config" #f #t sections: `( "default" ,target )))) ;; read the waivers config if it exists
;; (setup-env-defaults (conc *toppath* "/runconfigs.config") run-id (make-hash-table) keyvals target)
;; (set-run-config-vars run-id keyvals target) ;; (db:get-target db run-id))
;; Now have runconfigs data loaded, set environment vars
(for-each (lambda (section)
(for-each (lambda (varval)
(let ((var (car varval))
(val (cadr varval)))
(if (and (string? var)(string? val))
(begin
(setenv var (config:eval-string-in-environment val))) ;; val)
(debug:print-error 0 *default-log-port* "bad variable spec, " var "=" val))))
(configf:get-section rconfig section)))
(list "default" target)))
|
>
>
>
>
>
|
|
>
|
|
|
|
|
|
|
|
>
>
>
|
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
|
;; one of these is defunct/redundant ...
(if (not (launch:setup force-reread: #t))
(begin
(debug:print 0 *default-log-port* "Failed to setup, exiting")
;; (sqlite3:finalize! db)
;; (sqlite3:finalize! tdb)
(exit 1)))
;; validate that the test run area is available
(check-work-area)
;; still need to go back to run area home for next couple steps
(change-directory *toppath*)
;; NOTE: Current order is to process runconfigs *before* setting the MT_ vars. This
;; seems non-ideal but could well break stuff
;; BUG? BUG? BUG?
(let ((rconfig (full-runconfigs-read)) ;; (read-config (conc *toppath* "/runconfigs.config") #f #t sections: (list "default" target))))
(wconfig (read-config "waivers.config" #f #t sections: `( "default" ,target )))) ;; read the waivers config if it exists
;; (setup-env-defaults (conc *toppath* "/runconfigs.config") run-id (make-hash-table) keyvals target)
;; (set-run-config-vars run-id keyvals target) ;; (db:get-target db run-id))
;; Now have runconfigs data loaded, set environment vars
(for-each
(lambda (section)
(for-each
(lambda (varval)
(let ((var (car varval))
(val (cadr varval)))
(if (and (string? var)(string? val))
(begin
(setenv var (config:eval-string-in-environment val))) ;; val)
(debug:print-error 0 *default-log-port* "bad variable spec, " var "=" val))))
(configf:get-section rconfig section)))
(list "default" target)))
;;(bb-check-path msg: "launch:execute post block 1")
;; NFS might not have propagated the directory meta data to the run host - give it time if needed
(let loop ((count 0))
(if (or (common:file-exists? work-area)
(> count 10))
(change-directory work-area)
(begin
(debug:print 0 *default-log-port* "INFO: Not starting job yet - directory " work-area " not found")
(thread-sleep! 10)
(loop (+ count 1)))))
;; now we can switch to the work-area?
(change-directory work-area)
;;(bb-check-path msg: "launch:execute post block 1.5")
;; (change-directory work-area)
(set! keyvals (keys:target->keyval keys target))
;; apply pre-overrides before other variables. The pre-override vars must not
;; clobbers things from the official sources such as megatest.config and runconfigs.config
(if (string? set-vars)
(let ((varpairs (string-split set-vars ",")))
|
| ︙ | | | ︙ | |
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
|
(exit-info (make-launch:einf pid: #t exit-status: #t exit-code: #t rollup-status: 0)) ;; pid exit-status exit-code (i.e. process was successfully run) rollup-status
(job-thread #f)
;; (keep-going #t)
(misc-flags (let ((ht (make-hash-table)))
(hash-table-set! ht 'keep-going #t)
ht))
(runit (lambda ()
(launch:manage-steps run-id test-id item-path fullrunscript ezsteps test-name tconfigreg exit-info m)))
(monitorjob (lambda ()
(launch:monitor-job run-id test-id item-path fullrunscript ezsteps test-name tconfigreg exit-info m work-area runtlim misc-flags)))
(th1 (make-thread monitorjob "monitor job"))
(th2 (make-thread runit "run job")))
(set! job-thread th2)
(thread-start! th1)
(thread-start! th2)
|
|
|
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
|
(exit-info (make-launch:einf pid: #t exit-status: #t exit-code: #t rollup-status: 0)) ;; pid exit-status exit-code (i.e. process was successfully run) rollup-status
(job-thread #f)
;; (keep-going #t)
(misc-flags (let ((ht (make-hash-table)))
(hash-table-set! ht 'keep-going #t)
ht))
(runit (lambda ()
(launch:manage-steps run-id test-id item-path fullrunscript ezsteps subrun test-name tconfigreg exit-info m)))
(monitorjob (lambda ()
(launch:monitor-job run-id test-id item-path fullrunscript ezsteps test-name tconfigreg exit-info m work-area runtlim misc-flags)))
(th1 (make-thread monitorjob "monitor job"))
(th2 (make-thread runit "run job")))
(set! job-thread th2)
(thread-start! th1)
(thread-start! th2)
|
| ︙ | | | ︙ | |
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
|
(setenv (car kt) (cadr kt)))
key-vals)
(read-config (conc toppath "/runconfigs.config") *runconfigdat* #t ;; consider using runconfig:read some day ...
sections: sections)))
(cachefiles (launch:get-cache-file-paths areapath toppath target mtconfig))
(mtcachef (car cachefiles))
(rccachef (cdr cachefiles)))
(if rccachef (configf:write-alist runconfigdat rccachef))
(if mtcachef (configf:write-alist *configdat* mtcachef))
|
>
>
>
>
>
>
>
|
>
>
>
>
|
>
|
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
|
(setenv (car kt) (cadr kt)))
key-vals)
(read-config (conc toppath "/runconfigs.config") *runconfigdat* #t ;; consider using runconfig:read some day ...
sections: sections)))
(cachefiles (launch:get-cache-file-paths areapath toppath target mtconfig))
(mtcachef (car cachefiles))
(rccachef (cdr cachefiles)))
;; trap exception due to stale NFS handle -- Error: (open-output-file) cannot open file - Stale NFS file handle: "/p/fdk/gwa/lefkowit/mtTesting/qa/primbeqa/links/p1222/11/PDK_r1.1.1/prim/clean/pcell_testgen/.runconfigs.cfg-1.6427-7d1e789cb3f62f9cde719a4865bb51b3c17ea853" - ticket 220546342
;; TODO - consider 1) using simple-lock to bracket cache write
;; 2) cache in hash on server, since need to do rmt: anyway to lock.
(if rccachef
(common:fail-safe
(lambda ()
(configf:write-alist runconfigdat rccachef))
(conc "Could not write cache file - "rccachef)))
(if mtcachef
(common:fail-safe
(lambda ()
(configf:write-alist *configdat* mtcachef))
(conc "Could not write cache file - "mtcachef)))
(set! *runconfigdat* runconfigdat)
(if (and rccachef mtcachef) (set! *configstatus* 'fulldata))))
;; no configs found? should not happen but let's try to recover gracefully, return an empty hash-table
(set! *configdat* (make-hash-table))
)))
;; else read what you can and set the flag accordingly
|
| ︙ | | | ︙ | |
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
|
(set! *toppath* #f) ;; force it to be false so we return #f
#f))
;; one more attempt to cache the configs for future reading
(let* ((cachefiles (launch:get-cache-file-paths areapath toppath target mtconfig))
(mtcachef (car cachefiles))
(rccachef (cdr cachefiles)))
(if (and rccachef *runconfigdat* (not (common:file-exists? rccachef))) (configf:write-alist *runconfigdat* rccachef))
(if (and mtcachef *configdat* (not (common:file-exists? mtcachef))) (configf:write-alist *configdat* mtcachef))
(if (and rccachef mtcachef *runconfigdat* *configdat*)
(set! *configstatus* 'fulldata)))
;; if have -append-config then read and append here
(let ((cfname (args:get-arg "-append-config")))
(if (and cfname
(file-read-access? cfname))
|
>
>
>
>
|
>
>
>
>
>
|
>
>
>
>
>
|
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
|
(set! *toppath* #f) ;; force it to be false so we return #f
#f))
;; one more attempt to cache the configs for future reading
(let* ((cachefiles (launch:get-cache-file-paths areapath toppath target mtconfig))
(mtcachef (car cachefiles))
(rccachef (cdr cachefiles)))
;; trap exception due to stale NFS handle -- Error: (open-output-file) cannot open file - Stale NFS file handle: "...somepath.../.runconfigs.cfg-1.6427-7d1e789cb3f62f9cde719a4865bb51b3c17ea853" - ticket 220546342
;; TODO - consider 1) using simple-lock to bracket cache write
;; 2) cache in hash on server, since need to do rmt: anyway to lock.
(if (and rccachef *runconfigdat* (not (common:file-exists? rccachef)))
(common:fail-safe
(lambda ()
(configf:write-alist *runconfigdat* rccachef))
(conc "Could not write cache file - "rccachef))
)
(if (and mtcachef *configdat* (not (common:file-exists? mtcachef)))
(common:fail-safe
(lambda ()
(configf:write-alist *configdat* mtcachef))
(conc "Could not write cache file - "mtcachef))
)
(if (and rccachef mtcachef *runconfigdat* *configdat*)
(set! *configstatus* 'fulldata)))
;; if have -append-config then read and append here
(let ((cfname (args:get-arg "-append-config")))
(if (and cfname
(file-read-access? cfname))
|
| ︙ | | | ︙ | |
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
|
(let ((res (common:get-disk-with-most-free-space disks minspace))) ;; min size of 1000, seems tad dumb
(if res
(cdr res)
(begin
(if (common:low-noise-print 20 "No valid disks or no disk with enough space")
(debug:print-error 0 *default-log-port* "No valid disks found in megatest.config. Please add some to your [disks] section and ensure the directory exists and has enough space!\n You can change minspace in the [setup] section of megatest.config. Current setting is: " minspace))
(exit 1))))))) ;; TODO - move the exit to the calling location and return #f
;; Desired directory structure:
;;
;; <linkdir> - <target> - <testname> -.
;; |
;; v
;; <rundir> - <target> - <testname> -|- <itempath(s)>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
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
|
(let ((res (common:get-disk-with-most-free-space disks minspace))) ;; min size of 1000, seems tad dumb
(if res
(cdr res)
(begin
(if (common:low-noise-print 20 "No valid disks or no disk with enough space")
(debug:print-error 0 *default-log-port* "No valid disks found in megatest.config. Please add some to your [disks] section and ensure the directory exists and has enough space!\n You can change minspace in the [setup] section of megatest.config. Current setting is: " minspace))
(exit 1))))))) ;; TODO - move the exit to the calling location and return #f
(define (launch:test-copy test-src-path test-path)
(let* ((ovrcmd (let ((cmd (config-lookup *configdat* "setup" "testcopycmd")))
(if cmd
;; substitute the TEST_SRC_PATH and TEST_TARG_PATH
(string-substitute "TEST_TARG_PATH" test-path
(string-substitute "TEST_SRC_PATH" test-src-path cmd #t) #t)
#f)))
(cmd (if ovrcmd
ovrcmd
(conc "rsync -av" (if (debug:debug-mode 1) "" "q") " " test-src-path "/ " test-path "/"
" >> " test-path "/mt_launch.log 2>> " test-path "/mt_launch.log")))
(status (system cmd)))
(if (not (eq? status 0))
(debug:print 2 *default-log-port* "ERROR: problem with running \"" cmd "\""))))
;; Desired directory structure:
;;
;; <linkdir> - <target> - <testname> -.
;; |
;; v
;; <rundir> - <target> - <testname> -|- <itempath(s)>
|
| ︙ | | | ︙ | |
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
|
(if (not (common:file-exists? lnktarget)) (create-symbolic-link test-path lnktarget)))))
(if (not (directory? test-path))
(create-directory test-path #t)) ;; this is a hack, I don't know why out of the blue this path does not exist sometimes
(if (and test-src-path (directory? test-path))
(begin
(let* ((ovrcmd (let ((cmd (config-lookup *configdat* "setup" "testcopycmd")))
(if cmd
;; substitute the TEST_SRC_PATH and TEST_TARG_PATH
(string-substitute "TEST_TARG_PATH" test-path
(string-substitute "TEST_SRC_PATH" test-src-path cmd #t) #t)
#f)))
(cmd (if ovrcmd
ovrcmd
(conc "rsync -av" (if (debug:debug-mode 1) "" "q") " " test-src-path "/ " test-path "/"
" >> " test-path "/mt_launch.log 2>> " test-path "/mt_launch.log")))
(status (system cmd)))
(if (not (eq? status 0))
(debug:print 2 *default-log-port* "ERROR: problem with running \"" cmd "\"")))
(let* ((item-path (item-list->path itemdat))
|
<
<
<
<
|
<
<
<
<
<
<
<
<
>
>
>
>
>
>
>
>
>
>
>
>
>
|
|
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
|
(if (not (common:file-exists? lnktarget)) (create-symbolic-link test-path lnktarget)))))
(if (not (directory? test-path))
(create-directory test-path #t)) ;; this is a hack, I don't know why out of the blue this path does not exist sometimes
(if (and test-src-path (directory? test-path))
(begin
(launch:test-copy test-src-path test-path)
(list lnkpathf lnkpath ))
(if (and test-src-path (> remtries 0))
(begin
(debug:print-error 0 *default-log-port* "Failed to create work area at " test-path " with link at " lnktarget ", remaining attempts " remtries)
;;
(create-work-area run-id run-info keyvals test-id test-src-path disk-path testname itemdat remtries: (- remtries 1)))
(list #f #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: 1)))
(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"))
|
| ︙ | | | ︙ | |
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
|
(useshell (let ((ush (config-lookup *configdat* "jobtools" "useshell")))
(if ush
(if (equal? ush "no") ;; must use "no" to NOT use shell
#f
ush)
#t))) ;; default is yes
(runscript (config-lookup tconfig "setup" "runscript"))
(ezsteps (> (length (hash-table-ref/default tconfig "ezsteps" '())) 0)) ;; don't send all the steps, could be big
;; (diskspace (config-lookup tconfig "requirements" "diskspace"))
;; (memory (config-lookup tconfig "requirements" "memory"))
;; (hosts (config-lookup *configdat* "jobtools" "workhosts")) ;; I'm pretty sure this was never completed
(remote-megatest (config-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
|
|
>
|
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
|
(useshell (let ((ush (config-lookup *configdat* "jobtools" "useshell")))
(if ush
(if (equal? ush "no") ;; must use "no" to NOT use shell
#f
ush)
#t))) ;; default is yes
(runscript (config-lookup tconfig "setup" "runscript"))
(ezsteps (> (length (hash-table-ref/default tconfig "ezsteps" '())) 0)) ;; don't send all the steps, could be big, just send a flag
(subrun (> (length (hash-table-ref/default tconfig "subrun" '())) 0)) ;; send a flag to process a subrun
;; (diskspace (config-lookup tconfig "requirements" "diskspace"))
;; (memory (config-lookup tconfig "requirements" "memory"))
;; (hosts (config-lookup *configdat* "jobtools" "workhosts")) ;; I'm pretty sure this was never completed
(remote-megatest (config-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
|
| ︙ | | | ︙ | |
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
|
(list 'test-name test-name)
(list 'runscript runscript)
(list 'run-id run-id )
(list 'test-id test-id )
;; (list 'item-path item-path )
(list 'itemdat itemdat )
(list 'megatest remote-megatest)
(list 'ezsteps ezsteps)
(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))))))))
|
|
>
|
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
|
(list 'test-name test-name)
(list 'runscript runscript)
(list 'run-id run-id )
(list 'test-id test-id )
;; (list 'item-path item-path )
(list 'itemdat itemdat )
(list 'megatest remote-megatest)
(list 'ezsteps ezsteps)
(list 'subrun subrun)
(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))))))))
|
| ︙ | | | ︙ | |
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
|
cmdstr
(conc cmdstr " >> mt_launch.log 2>&1 &")))
(car fullcmd))
(if useshell
'()
(cdr fullcmd)))))
(mutex-unlock! *launch-setup-mutex*) ;; yes, really should mutex all the way to here. Need to put this entire process into a fork.
(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)
(apply print launch-results)
|
>
|
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
|
cmdstr
(conc cmdstr " >> mt_launch.log 2>&1 &")))
(car fullcmd))
(if useshell
'()
(cdr fullcmd)))))
(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)
(apply print launch-results)
|
| ︙ | | | ︙ | |