| ︙ | | |
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
|
19
20
21
22
23
24
25
26
27
28
29
30
31
32
|
-
-
-
|
(import (prefix base64 base64:))
(import (prefix sqlite3 sqlite3:))
(declare (unit launch))
(declare (uses common))
(declare (uses configf))
(declare (uses db))
;; (declare (uses sdb))
(declare (uses tdb))
;; (declare (uses filedb))
|
| ︙ | | |
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
|
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
|
-
+
+
+
+
+
-
-
+
+
+
+
+
+
-
+
+
+
+
-
+
-
+
+
+
-
-
+
+
+
+
+
+
+
-
+
|
(defstruct launch:einf (pid #t)(exit-status #t)(exit-code #t)(rollup-status 0))
;; return (conc status ": " comment) from the final section so that
;; the comment can be set in the step record in launch.scm
;;
(define (launch:load-logpro-dat run-id test-id stepname)
(let ((cname (conc stepname ".dat")))
(if (file-exists? cname)
(if (common:file-exists? cname)
(let* ((dat (read-config cname #f #f))
(csvr (db:logpro-dat->csv dat stepname))
(csvt (let-values (((fmt-cell fmt-record fmt-csv) (make-format ",")))
(fmt-csv (map list->csv-record csvr))))
(status (configf:lookup dat "final" "exit-status"))
(msg (configf:lookup dat "final" "message")))
(if csvt ;; this if blocked stack dump caused by .dat file from logpro being 0-byte. fixed by upgrading logpro
(rmt:csv->test-data run-id test-id csvt)
(debug:print 0 *default-log-port* "ERROR: no csvdat exists for run-id: " run-id " test-id: " test-id " stepname: " stepname ", check that logpro version is 1.15 or newer"))
;; (debug:print-info 13 *default-log-port* "Error: run-id/test-id/stepname="run-id"/"test-id"/"stepname" => bad csvr="csvr)
;; )
(cond
((equal? status "PASS") "PASS") ;; skip the message part if status is pass
(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*|)(.*)$") stepinfo))
(stepparms (list-ref stepparts 2)) ;; for future use, {VAR=1,2,3}, run step for each
(stepparts (string-match (regexp "^(\\{([^\\}\\{]*)\\}\\s*|)(.*)$") 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 (file-exists? logpro-file)))
(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
" stepparms: " stepparms " stepcmd: " stepcmd)
" 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 (file-exists? prev-env))
;; (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 ()
(pid (process-run "/bin/bash" (list "-c" cmd))))
(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 (file-exists? (conc stepname ".logpro"))
(if (common:file-exists? (conc stepname ".logpro"))
(print "\tlogpro " stepname ".logpro " stepname ".html < " stepname ".log"))
(print)
(print stepname " : " stepname ".log")
(print))
#:append)
(rmt:test-set-top-process-pid run-id test-id pid)
|
| ︙ | | |
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
|
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
|
-
+
|
(let ((exinfo (launch:einf-exit-code exit-info)) ;; (vector-ref exit-info 2))
(logfna (if logpro-used (conc stepname ".html") ""))
(comment #f))
(if logpro-used
(let ((datfile (conc stepname ".dat")))
;; load the .dat file into the test_data table if it exists
(if (file-exists? datfile)
(if (common:file-exists? datfile)
(set! comment (launch:load-logpro-dat run-id test-id stepname)))
(rmt:test-set-log! run-id test-id (conc stepname ".html"))))
(rmt:teststep-set-status! run-id test-id stepname "end" exinfo comment logfna))
;; set the test final status
(let* ((process-exit-status (launch:einf-exit-code exit-info)) ;; (vector-ref exit-info 2))
(this-step-status (cond
((and (eq? process-exit-status 2) logpro-used) 'warn) ;; logpro 2 = warnings
|
| ︙ | | |
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
|
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
|
-
+
|
(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)
(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
|
| ︙ | | |
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
|
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
|
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
+
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
+
|
(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
(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 (let ((ra (configf:lookup testconfig "subrun" "runarea")))
(if ra ;; when runarea is not set we default to *toppath*. However
ra ;; we need to force the setting in the testconfig so it will
(begin ;; be preserved in the testconfig.subrun file
(configf:set-section-var testconfig "subrun" "runarea" *toppath*)
*toppath*))))
(passfail (configf:lookup testconfig "subrun" "passfail"))
(target (or (configf:lookup testconfig "subrun" "target") (get-environment-variable "MT_TARGET")))
(runname (or (configf:lookup testconfig "subrun" "runname")(get-environment-variable "MT_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 " runname
(conc " -start-dir " runarea) ;; (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 (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 (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)))))))
(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")))
(let* ((start-seconds (current-seconds))
(start-seconds (current-seconds))
(calc-minutes (lambda ()
(inexact->exact
(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))))
(let ((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 (get-df (current-directory)))
(delta (abs (- df disk-free))))
(if (> delta 200) ;; ignore changes under 200 Meg
df
#f))))
(disk-free (get-df (current-directory)))
(last-sync (current-seconds)))
(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
(get-df (current-directory))
disk-free))
(delta (abs (- df disk-free))))
(if (and (> df 0)
(> (/ delta df) 0.1)) ;; (> delta 200) ;; ignore changes under 200 Meg
df
#f)))
(do-sync (or new-cpu-load new-disk-free over-time)))
(debug:print 4 *default-log-port* "cpu: " new-cpu-load " disk: " new-disk-free " last-sync: " last-sync " do-sync: " do-sync)
(set! kill-job? (or (test-get-kill-request run-id test-id) ;; run-id test-name itemdat))
(and runtlim (let* ((run-seconds (- (current-seconds) start-seconds))
(time-exceeded (> run-seconds runtlim)))
(if time-exceeded
(begin
(debug:print-info 0 *default-log-port* "KILLING TEST DUE TO TIME LIMIT EXCEEDED! Runtime=" run-seconds " seconds, limit=" runtlim)
#t)
#f)))))
(if do-sync
(tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f)
|
| ︙ | | |
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
|
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
|
-
+
+
+
+
+
|
(mutex-unlock! m)
;; no point in sticking around. Exit now.
(exit)))
(if (hash-table-ref/default misc-flags 'keep-going #f)
(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)))))))
(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")
(if (list? cmdinfo) ;; ((testpath /tmp/mrwellan/jazzmind/src/example_run/tests/sqlitespeed)
;; (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))
|
| ︙ | | |
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
|
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
|
-
-
+
+
+
+
+
+
+
+
+
+
+
+
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
-
-
-
-
-
-
-
-
-
-
-
+
|
(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 (file-exists? fulln)
(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)))))
) ;; (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 (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))
|
| ︙ | | |
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
|
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
|
+
-
+
+
+
+
+
-
+
|
(set-signal-handler! signal/term sighand)
) ;; (set-signal-handler! signal/stop sighand)
;; Do not run the test if it is REMOVING, RUNNING, KILLREQ or REMOTEHOSTSTART,
;; Mark the test as REMOTEHOSTSTART *IMMEDIATELY*
;;
(let* ((test-info (rmt:get-test-info-by-id run-id test-id))
(test-host (if test-info
(test-host (db:test-get-host 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" "KILLREQ" "STUCK")) ;; prior run of this test didn't complete, go ahead and try to rerun
((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")
|
| ︙ | | |
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
|
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
|
-
+
+
+
+
+
+
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
-
+
+
+
+
|
(debug:print-error 0 *default-log-port* "test state is " (db:test-get-state test-info) ", cannot proceed")
(exit))))
(debug:print 2 *default-log-port* "Exectuing " test-name " (id: " test-id ") on " (get-host-name))
(set! keys (rmt:get-keys))
;; (runs:set-megatest-env-vars run-id inkeys: keys inkeyvals: keyvals) ;; these may be needed by the launching process
;; one of these is defunct/redundant ...
(if (not (launch:setup force: #t))
(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
(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)))
(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 (file-exists? work-area)
(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 ",")))
|
| ︙ | | |
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
|
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
|
+
+
+
-
+
-
+
+
+
+
+
+
+
-
+
|
;; environment overrides are done *before* the remaining critical envars.
(alist->env-vars env-ovrd)
;;(bb-check-path msg: "launch:execute post block 41")
(runs:set-megatest-env-vars run-id inkeys: keys inkeyvals: keyvals)
;;(bb-check-path msg: "launch:execute post block 42")
(set-item-env-vars itemdat)
;;(bb-check-path msg: "launch:execute post block 43")
(let ((blacklist (configf:lookup *configdat* "setup" "blacklistvars")))
(if blacklist
(save-environment-as-files "megatest" ignorevars: (string-split blacklist))
(save-environment-as-files "megatest")
(file-exists? fullrunscript)
(common:file-exists? fullrunscript)
(not (file-execute-access? fullrunscript)))
(system (conc "chmod ug+x " fullrunscript))))
;; We are about to actually kick off the test
;; so this is a good place to remove the records for
;; any previous runs
;; (db:test-remove-steps db run-id testname itemdat)
;; now is also a good time to write the .testconfig file
(let* ((tconfig-fname (conc work-area "/.testconfig"))
(tconfig-tmpfile (conc tconfig-fname ".tmp"))
(tconfig (tests:get-testconfig test-name item-path tconfigreg #t force-create: #t))) ;; 'return-procs)))
(configf:write-alist tconfig tconfig-tmpfile)
(file-move tconfig-tmpfile tconfig-fname #t))
;;
(let* ((m (make-mutex))
(kill-job? #f)
(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)))
(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)
|
| ︙ | | |
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
|
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
|
-
+
+
+
+
-
+
-
+
-
+
-
+
|
(tests:summarize-items run-id test-id test-name #f))
(tests:summarize-test run-id test-id) ;; don't force - just update if no
(rmt:update-run-stats run-id (rmt:get-raw-run-stats run-id)))
(mutex-unlock! m)
(debug:print 2 *default-log-port* "Output from running " fullrunscript ", pid " (launch:einf-pid exit-info) " in work area "
work-area ":\n====\n exit code " (launch:einf-exit-code exit-info) "\n" "====\n")
(if (not (launch:einf-exit-status exit-info))
(exit 4)))))))
(exit 4))))
)))
;; DO NOT USE - caching of configs is handled in launch:setup now.
;;
(define (launch:cache-config)
;; if we have a linktree and -runtests and -target and the directory exists dump the config
;; to megatest-(current-seconds).cfg and symlink it to megatest.cfg
(if (and *configdat*
(or (args:get-arg "-run")
(args:get-arg "-runtests")
(args:get-arg "-execute")))
(let* ((linktree (common:get-linktree)) ;; (get-environment-variable "MT_LINKTREE"))
(target (common:args-get-target exit-if-bad: #t))
(runname (or (args:get-arg "-runname")
(args:get-arg ":runname")
(getenv "MT_RUNNAME")))
(fulldir (conc linktree "/"
target "/"
runname)))
(if (and linktree (file-exists? linktree)) ;; can't proceed without linktree
(if (and linktree (common:file-exists? linktree)) ;; can't proceed without linktree
(begin
(debug:print-info 0 *default-log-port* "Have -run with target=" target ", runname=" runname ", fulldir=" fulldir ", testpatt=" (or (args:get-arg "-testpatt") "%"))
(if (not (file-exists? fulldir))
(if (not (common:file-exists? fulldir))
(create-directory fulldir #t)) ;; need to protect with exception handler
(if (and target
runname
(file-exists? fulldir))
(common:file-exists? fulldir))
(let ((tmpfile (conc fulldir "/.megatest.cfg." (current-seconds)))
(targfile (conc fulldir "/.megatest.cfg-" megatest-version "-" megatest-fossil-hash))
(rconfig (conc fulldir "/.runconfig." megatest-version "-" megatest-fossil-hash)))
(if (file-exists? rconfig) ;; only cache megatest.config AFTER runconfigs has been cached
(if (common:file-exists? rconfig) ;; only cache megatest.config AFTER runconfigs has been cached
(begin
(debug:print-info 0 *default-log-port* "Caching megatest.config in " tmpfile)
(if (not (common:in-running-test?))
(configf:write-alist *configdat* tmpfile))
(system (conc "ln -sf " tmpfile " " targfile))))
)))
(debug:print-info 1 *default-log-port* "No linktree yet, no caching configs.")))))
|
| ︙ | | |
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
|
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
|
-
+
-
+
-
+
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
+
-
-
-
-
+
+
-
+
-
-
+
-
-
-
+
+
+
+
-
-
-
+
-
-
-
-
+
+
+
+
-
-
+
+
+
+
+
+
+
+
-
+
+
+
+
+
|
;; returns:
;; *toppath*
;; side effects:
;; sets; *configdat* (megatest.config info)
;; *runconfigdat* (runconfigs.config info)
;; *configstatus* (status of the read data)
;;
(define (launch:setup #!key (force #f) (areapath #f))
(define (launch:setup #!key (force-reread #f) (areapath #f))
(mutex-lock! *launch-setup-mutex*)
(if (and *toppath*
(eq? *configstatus* 'fulldata)) ;; got it all
(eq? *configstatus* 'fulldata) (not force-reread)) ;; got it all
(begin
(debug:print 0 *default-log-port* "NOTE: skipping launch:setup-body call since we have fulldata")
(debug:print 2 *default-log-port* "NOTE: skipping launch:setup-body call since we have fulldata")
(mutex-unlock! *launch-setup-mutex*)
*toppath*)
(let ((res (launch:setup-body force: force areapath: areapath)))
(let ((res (launch:setup-body force-reread: force-reread areapath: areapath)))
(mutex-unlock! *launch-setup-mutex*)
res)))
;; return paths depending on what info is available.
;;
(define (launch:get-cache-file-paths areapath toppath target mtconfig)
(let* ((use-cache (common:use-cache?))
(runname (common:args-get-runname))
(linktree (common:get-linktree))
(testname (common:get-full-test-name))
(rundir (if (and runname target linktree)
(common:directory-writable? (conc linktree "/" target "/" runname))
#f))
(testdir (if (and rundir testname)
(common:directory-writable? (conc rundir "/" testname))
#f))
(cachedir (or testdir rundir))
(mtcachef (and cachedir (conc cachedir "/" ".megatest.cfg-" megatest-version "-" megatest-fossil-hash)))
(rccachef (and cachedir (conc cachedir "/" ".runconfigs.cfg-" megatest-version "-" megatest-fossil-hash))))
(debug:print-info 6 *default-log-port*
"runname=" runname
"\n linktree=" linktree
"\n testname=" testname
"\n rundir=" rundir
"\n testdir=" testdir
"\n cachedir=" cachedir
"\n mtcachef=" mtcachef
"\n rccachef=" rccachef)
(cons mtcachef rccachef)))
(define (launch:setup-body #!key (force-reread #f) (areapath #f))
(if (and (eq? *configstatus* 'fulldata)
*toppath*
(not force-reread)) ;; no need to reprocess
*toppath* ;; return toppath
(let* ((use-cache (common:use-cache?))
(let* ((use-cache (common:use-cache?)) ;; BB- use-cache checks *configdat* for use-cache setting. We do not have *configdat*. Bootstrapping problem here.
(toppath (or *toppath* areapath (getenv "MT_RUN_AREA_HOME"))) ;; preserve toppath
(runname (common:args-get-runname))
(linktree (common:get-linktree))
(contour #f) ;; NOT READY FOR THIS (args:get-arg "-contour"))
(rundir (if (and runname target linktree)
(mtcachef (if (null? cachefiles)
(conc linktree (if contour (conc "/" contour) "") "/" target "/" runname)
#f))
(mtcachef (and rundir (conc rundir "/" ".megatest.cfg-" megatest-version "-" megatest-fossil-hash)))
(rccachef (and rundir (conc rundir "/" ".runconfigs.cfg-" megatest-version "-" megatest-fossil-hash)))
(cancreate (and rundir (common:file-exists? rundir)(file-write-access? rundir) (not (common:in-running-test?)))))
;; (cxt (hash-table-ref/default *contexts* toppath #f)))
;; create our cxt for this area if it doesn't already exist
;; (if (not cxt)(hash-table-set! *contexts* toppath (make-cxt)))
;; (print "runname: " runname " target: " target " mtcachef: " mtcachef " rccachef: " rccachef)
((and mtcachef (common:file-exists? mtcachef) (get-environment-variable "MT_RUN_AREA_HOME") use-cache)
(set! *configdat* (configf:read-alist mtcachef))
(get-environment-variable "MT_RUN_AREA_HOME")
(common:file-exists? mtcachef)
(common:file-exists? rccachef))
;;(BB> "launch:setup-body -- cond branch 1 - use-cache")
(set! *configdat* (configf:read-alist mtcachef))
;;(BB> "launch:setup-body -- 1 set! *configdat*="*configdat*)
(set! *runconfigdat* (configf:read-alist rccachef))
(set! *configinfo* (list *configdat* (get-environment-variable "MT_RUN_AREA_HOME")))
(set! *configstatus* 'fulldata)
(set! *toppath* (get-environment-variable "MT_RUN_AREA_HOME"))
*toppath*)
;; there are no existing cached configs, do full reads of the configs and cache them
;; we have all the info needed to fully process runconfigs and megatest.config
((and ;; (not force-reread) ;; force-reread is irrelevant in the AND, could however OR it?
(mtcachef
mtcachef
rccachef) ;; BB- why are we doing this without asking if caching is desired?
;;(BB> "launch:setup-body -- cond branch 2")
(let* ((first-pass (find-and-read-config ;; NB// sets MT_RUN_AREA_HOME as side effect
mtconfig
environ-patt: "env-override"
given-toppath: toppath
pathenvvar: "MT_RUN_AREA_HOME"))
(first-rundat (let ((toppath (if toppath
toppath
(car first-pass))))
(read-config ;; (conc toppath "/runconfigs.config") ;; this should be converted to runconfig:read but it is non-trivial, leaving it for now.
(conc (if (string? toppath)
toppath
(get-environment-variable "MT_RUN_AREA_HOME"))
"/runconfigs.config")
*runconfigdat* #t
sections: sections))))
(set! *runconfigdat* first-rundat)
(if first-pass ;;
(begin
;;(BB> "launch:setup-body -- \"first-pass\"=first-pass")
(set! *configdat* (car first-pass))
;;(BB> "launch:setup-body -- 2 set! *configdat*="*configdat*)
(set! *configinfo* first-pass)
(set! *toppath* (or toppath (cadr first-pass))) ;; use the gathered data unless already have it
(set! toppath *toppath*)
(if (not *toppath*)
(begin
(debug:print-error 0 *default-log-port* "you are not in a megatest area!")
(exit 1)))
|
| ︙ | | |
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
|
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
|
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
+
+
+
+
-
+
+
+
+
-
+
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
given-toppath: toppath
pathenvvar: "MT_RUN_AREA_HOME"))
(runconfigdat (begin ;; this read of the runconfigs will see any adjustments made by re-reading megatest.config
(for-each (lambda (kt)
(setenv (car kt) (cadr kt)))
key-vals)
(read-config (conc toppath "/runconfigs.config") *runconfigdat* #t ;; consider using runconfig:read some day ...
sections: sections))))
(if cancreate (configf:write-alist runconfigdat rccachef))
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 cancreate (configf:write-alist *configdat* mtcachef))
(if cancreate (set! *configstatus* 'fulldata))))
(if cfgdat
(if (not (file-exists? tlink))
(if (not (common:file-exists? tlink))
(create-symbolic-link linktree tlink)))))
(begin
(debug:print-error 0 *default-log-port* "linktree not defined in [setup] section of megatest.config")
)))
(if (and *toppath*
(directory-exists? *toppath*))
(begin
(setenv "MT_RUN_AREA_HOME" *toppath*)
(setenv "MT_TESTSUITENAME" (common:get-testsuite-name)))
(begin
(debug:print-error 0 *default-log-port* "failed to find the top path to your Megatest area.")
;;(exit 1)
#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))
(read-config cfname *configdat* #t))) ;; values are added to the hash, no need to do anything special.
*toppath*)))
(define (get-best-disk confdat testconfig)
(let* ((disks (or (and testconfig (hash-table-ref/default testconfig "disks" #f))
(hash-table-ref/default confdat "disks" #f)))
(minspace (let ((m (configf:lookup confdat "setup" "minspace")))
(string->number (or m "10000")))))
(if disks
(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)>
|
| ︙ | | |
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
|
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
|
-
+
|
(handle-exceptions
exn
(begin
(debug:print-error 0 *default-log-port* " Failed to remove symlink " lnkpath ((condition-property-accessor 'exn 'message) exn) ", exiting")
(exit 1))
(delete-file lnkpath)))
(if (not (or (file-exists? lnkpath)
(if (not (or (common:file-exists? lnkpath)
(symbolic-link? lnkpath)))
(handle-exceptions
exn
(begin
(debug:print-error 0 *default-log-port* " Failed to create symlink " lnkpath ((condition-property-accessor 'exn 'message) exn) ", exiting")
(exit 1))
(create-symbolic-link toptest-path lnkpath)))
|
| ︙ | | |
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
|
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
|
-
+
|
;; (db:get-path dbstruct
;; (rmt:sdb-qry 'getstr
(db:test-get-rundir testinfo) ;; ) ;; )
#f)))
(hash-table-set! *toptest-paths* testname curr-test-path)
;; NB// Was this for the test or for the parent in an iterated test?
(rmt:general-call 'test-set-rundir-shortdir run-id lnkpath
(if (file-exists? lnkpath)
(if (common:file-exists? lnkpath)
;; (resolve-pathname lnkpath)
(common:nice-path lnkpath)
lnkpath)
testname "" run-id)
;; (rmt:general-call 'test-set-rundir run-id lnkpath testname "") ;; toptest-path)
(if (or (not curr-test-path)
(not (directory-exists? toptest-path)))
|
| ︙ | | |
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
|
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
|
-
+
-
-
-
-
-
+
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
-
+
-
+
+
-
+
|
;; If there is already a symlink delete it and recreate it.
(handle-exceptions
exn
(begin
(debug:print-error 0 *default-log-port* " 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)))))
(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))
(launch-delay (string->number (or (configf:lookup *configdat* "setup" "launch-delay") "5"))))
(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* "Delaying launch of " test-name " for " (- launch-delay delta) " seconds")
(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*)
|
| ︙ | | |
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
|
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
|
-
+
+
|
(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
(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
|
| ︙ | | |
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
|
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
|
-
+
+
-
+
|
(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 '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))))))))
;; 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 (file-exists? work-area)
(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)))
|
| ︙ | | |
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
|
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
|
+
|
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)
|
| ︙ | | |