Megatest

Diff
Login

Differences From Artifact [b78a28ba5c]:

To Artifact [3c2150da95]:


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
			"-itempatt"
			"-setlog"
			"-set-toplog"
			"-runstep"
			"-logpro"
			"-m"
			"-rerun"

			) 
		 (list  "-h"
		        "-force"
		        "-xterm"
		        "-showkeys"
		        "-test-status"
		        "-gui"
			"-runall"    ;; run all tests
			"-remove-runs"
			"-keepgoing"
			"-usequeue"


		       )
		 args:arg-hash
		 0))

(if (args:get-arg "-h")
    (begin
      (print help)







>











>
>







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
			"-itempatt"
			"-setlog"
			"-set-toplog"
			"-runstep"
			"-logpro"
			"-m"
			"-rerun"
			"-debug" ;; for *verbosity* > 2
			) 
		 (list  "-h"
		        "-force"
		        "-xterm"
		        "-showkeys"
		        "-test-status"
		        "-gui"
			"-runall"    ;; run all tests
			"-remove-runs"
			"-keepgoing"
			"-usequeue"
			"-v" ;; verbose 2, more than normal (normal is 1)
			"-q" ;; quiet 0, errors/warnings only
		       )
		 args:arg-hash
		 0))

(if (args:get-arg "-h")
    (begin
      (print help)
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
156
157
158

(define *didsomething* #f)

;;======================================================================
;; Misc setup stuff
;;======================================================================







;;======================================================================
;; Remove old run(s)
;;======================================================================

;; since several actions can be specified on the command line the removal
;; is done first
(define (remove-runs)
  (cond
   ((not (args:get-arg ":runname"))
    (print "ERROR: Missing required parameter for -remove-runs, you must specify the run name pattern with :runname patt")
    (exit 2))
   ((not (args:get-arg "-testpatt"))
    (print "ERROR: Missing required parameter for -remove-runs, you must specify the test pattern with -testpatt")
    (exit 3))
   ((not (args:get-arg "-itempatt"))
    (print "ERROR: Missing required parameter for -remove-runs, you must specify the items with -itempatt")
    (exit 4))
   ((let ((db #f))
      (if (not (setup-for-run))
	  (begin 
	    (print "Failed to setup, exiting")
	    (exit 1)))
      (set! db (open-db))
      (if (not (car *configinfo*))
	  (begin
	    (print "ERROR: Attempted to remove test(s) but run area config file not found")
	    (exit 1))
	  ;; put test parameters into convenient variables
	  (runs:remove-runs db
			    (args:get-arg ":runname")
			    (args:get-arg "-testpatt")
			    (args:get-arg "-itempatt")))
      (sqlite3:finalize! db)







>
>
>
>
>
>









|


|







|




|







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
156
157
158
159
160
161
162
163
164
165
166
167

(define *didsomething* #f)

;;======================================================================
;; Misc setup stuff
;;======================================================================

(define *verbosity* (cond
		     ((args:get-arg "-debug")(string->number (args:get-arg "-debug")))
		     ((args:get-arg "-v")    2)
		     ((args:get-arg "-q")    0)
		     (else                   1)))

;;======================================================================
;; Remove old run(s)
;;======================================================================

;; since several actions can be specified on the command line the removal
;; is done first
(define (remove-runs)
  (cond
   ((not (args:get-arg ":runname"))
    (debug:print 0 "ERROR: Missing required parameter for -remove-runs, you must specify the run name pattern with :runname patt")
    (exit 2))
   ((not (args:get-arg "-testpatt"))
    (debug:print 0 "ERROR: Missing required parameter for -remove-runs, you must specify the test pattern with -testpatt")
    (exit 3))
   ((not (args:get-arg "-itempatt"))
    (print "ERROR: Missing required parameter for -remove-runs, you must specify the items with -itempatt")
    (exit 4))
   ((let ((db #f))
      (if (not (setup-for-run))
	  (begin 
	    (debug:print 0 print "Failed to setup, exiting")
	    (exit 1)))
      (set! db (open-db))
      (if (not (car *configinfo*))
	  (begin
	    (debug:print 0 "ERROR: Attempted to remove test(s) but run area config file not found")
	    (exit 1))
	  ;; put test parameters into convenient variables
	  (runs:remove-runs db
			    (args:get-arg ":runname")
			    (args:get-arg "-testpatt")
			    (args:get-arg "-itempatt")))
      (sqlite3:finalize! db)
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
	   (runs     (db:get-rows runsdat))
	   (header   (db:get-header runsdat))
	   (keys     (db-get-keys db))
	   (keynames (map key:get-fieldname keys)))
      ;; Each run
      (for-each 
       (lambda (run)
	 (print "Run: "
		(string-intersperse (map (lambda (x)
					   (db:get-value-by-header run header x))
					 keynames) "/")
		"/"
		(db:get-value-by-header run header "runname"))
	 (let ((run-id (db:get-value-by-header run header "id")))
	   (let ((tests (db-get-tests-for-run db run-id testpatt itempatt)))







|







185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
	   (runs     (db:get-rows runsdat))
	   (header   (db:get-header runsdat))
	   (keys     (db-get-keys db))
	   (keynames (map key:get-fieldname keys)))
      ;; Each run
      (for-each 
       (lambda (run)
	 (debug:print 2 "Run: "
		(string-intersperse (map (lambda (x)
					   (db:get-value-by-header run header x))
					 keynames) "/")
		"/"
		(db:get-value-by-header run header "runname"))
	 (let ((run-id (db:get-value-by-header run header "id")))
	   (let ((tests (db-get-tests-for-run db run-id testpatt itempatt)))
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
			 (lambda (step)
			   (format #t 
				   "    Step: ~20a State: ~10a Status: ~10a Time ~22a\n"
				   (db:step-get-stepname step)
				   (db:step-get-state step)
				   (db:step-get-status step)
				   (db:step-get-event_time step)))
			;;    (print "    Step: " (db:step-get-stepname step)
			;; 	  " " (db:step-get-state step)
			;; 	  " " (db:step-get-status step)
			;; 	  " " (db:step-get-event_time step)))
			 steps)))))
		tests))))
       runs)
      (set! *didsomething* #t)
      ))

;;======================================================================







<
<
<
<







226
227
228
229
230
231
232




233
234
235
236
237
238
239
			 (lambda (step)
			   (format #t 
				   "    Step: ~20a State: ~10a Status: ~10a Time ~22a\n"
				   (db:step-get-stepname step)
				   (db:step-get-state step)
				   (db:step-get-status step)
				   (db:step-get-event_time step)))




			 steps)))))
		tests))))
       runs)
      (set! *didsomething* #t)
      ))

;;======================================================================
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
;; if still ok to run tasks
;;   process deferred tasks per above steps

;; run all tests are are Not COMPLETED and PASS or CHECK
(if (args:get-arg "-runall")
    (if (not (args:get-arg ":runname"))
	(begin
	  (print "ERROR: Missing required parameter for -runtests, you must specify the run name with :runname runname")
	  (exit 2))
	(let* ((db      (if (setup-for-run)
			    (open-db)
			    (begin
			      (print "Failed to setup, exiting")
			      (exit 1)))))
	  (if (not (car *configinfo*))
	      (begin
		(print "ERROR: Attempted to run a test but run area config file not found")
		(exit 1))
	      ;; put test parameters into convenient variables
	      (let* ((test-names (get-all-legal-tests))) ;; "PROD" is ignored for now
		(print "INFO: Attempting to start the following tests...")
		(print "     " (string-intersperse test-names ","))
		(run-tests db test-names)))
	  ;; (run-waiting-tests db)
	  (sqlite3:finalize! db)
	  (set! *didsomething* #t))))

;;======================================================================
;; run one test







|




|



|



|
|







254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
;; if still ok to run tasks
;;   process deferred tasks per above steps

;; run all tests are are Not COMPLETED and PASS or CHECK
(if (args:get-arg "-runall")
    (if (not (args:get-arg ":runname"))
	(begin
	  (debug:print 0 "ERROR: Missing required parameter for -runtests, you must specify the run name with :runname runname")
	  (exit 2))
	(let* ((db      (if (setup-for-run)
			    (open-db)
			    (begin
			      (debug:print 0 "Failed to setup, exiting")
			      (exit 1)))))
	  (if (not (car *configinfo*))
	      (begin
		(debug:print 0 "ERROR: Attempted to run a test but run area config file not found")
		(exit 1))
	      ;; put test parameters into convenient variables
	      (let* ((test-names (get-all-legal-tests))) ;; "PROD" is ignored for now
		(debug:print 1 "INFO: Attempting to start the following tests...")
		(debug:print 1 "     " (string-intersperse test-names ","))
		(run-tests db test-names)))
	  ;; (run-waiting-tests db)
	  (sqlite3:finalize! db)
	  (set! *didsomething* #t))))

;;======================================================================
;; run one test
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
;; 6. test phone home
;;    - if test run time > allowed run time then kill job
;;    - if cannot access db > allowed disconnect time then kill job

(define (runtests)
  (if (not (args:get-arg ":runname"))
      (begin
	(print "ERROR: Missing required parameter for -runtests, you must specify the run name with :runname runname")
	(exit 2))
      (let ((db #f))
	(if (not (setup-for-run))
	    (begin 
	      (print "Failed to setup, exiting")
	      (exit 1)))
	(set! db (open-db))
	(if (not (car *configinfo*))
	    (begin
	      (print "ERROR: Attempted to run a test but run area config file not found")
	      (exit 1))
	    ;; put test parameters into convenient variables
	    (let* ((test-names   (string-split (args:get-arg "-runtests") ",")))
	      (run-tests db test-names)))
	;; run-waiting-tests db)
	(sqlite3:finalize! db)
	;; (run-waiting-tests #f)







|




|




|







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
;; 6. test phone home
;;    - if test run time > allowed run time then kill job
;;    - if cannot access db > allowed disconnect time then kill job

(define (runtests)
  (if (not (args:get-arg ":runname"))
      (begin
	(debug:print 0 "ERROR: Missing required parameter for -runtests, you must specify the run name with :runname runname")
	(exit 2))
      (let ((db #f))
	(if (not (setup-for-run))
	    (begin 
	      (debug:print 0 "Failed to setup, exiting")
	      (exit 1)))
	(set! db (open-db))
	(if (not (car *configinfo*))
	    (begin
	      (debug:print 0 "ERROR: Attempted to run a test but run area config file not found")
	      (exit 1))
	    ;; put test parameters into convenient variables
	    (let* ((test-names   (string-split (args:get-arg "-runtests") ",")))
	      (run-tests db test-names)))
	;; run-waiting-tests db)
	(sqlite3:finalize! db)
	;; (run-waiting-tests #f)
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
		 (itemdat   (assoc/default 'itemdat   cmdinfo))
		 (env-ovrd  (assoc/default 'env-ovrd  cmdinfo))
		 (runname   (assoc/default 'runname   cmdinfo))
		 (megatest  (assoc/default 'megatest  cmdinfo))
		 (mt-bindir-path (assoc/default 'mt-bindir-path cmdinfo))
		 (fullrunscript (conc testpath "/" runscript))
		 (db        #f))
	    (print "Exectuing " test-name " on " (get-host-name))
	    (change-directory testpath)
	    (setenv "MT_TEST_RUN_DIR" work-area)
	    (setenv "MT_TEST_NAME" test-name)
	    (setenv "MT_ITEM_INFO" (conc itemdat))
	    (setenv "MT_RUNNAME"   runname)
	    (setenv "MT_MEGATEST"  megatest)
	    (setenv "PATH" (conc (getenv "PATH") ":" mt-bindir-path))
	    
	    (if (not (setup-for-run))
		(begin
		  (print "Failed to setup, exiting") 
		  (exit 1)))
	    ;; now can find our db
	    (set! db (open-db))
	    (change-directory work-area) 
	    (let ((runconfigf (conc  *toppath* "/runconfigs.config")))
	      (if (file-exists? runconfigf)
		  (setup-env-defaults db runconfigf run-id)
		  (print "WARNING: You do not have a run config file: " runconfigf)))
            ;; environment overrides are done *before* the remaining critical envars.
            (alist->env-vars env-ovrd)
	    (set-megatest-env-vars db run-id)
	    (set-item-env-vars itemdat)
            (save-environment-as-files "megatest")
	    (test-set-meta-info db run-id test-name itemdat)
	    (test-set-status! db run-id test-name "REMOTEHOSTSTART" "n/a" itemdat (args:get-arg "-m"))







|










|







|







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
		 (itemdat   (assoc/default 'itemdat   cmdinfo))
		 (env-ovrd  (assoc/default 'env-ovrd  cmdinfo))
		 (runname   (assoc/default 'runname   cmdinfo))
		 (megatest  (assoc/default 'megatest  cmdinfo))
		 (mt-bindir-path (assoc/default 'mt-bindir-path cmdinfo))
		 (fullrunscript (conc testpath "/" runscript))
		 (db        #f))
	    (debug:print 2 "Exectuing " test-name " on " (get-host-name))
	    (change-directory testpath)
	    (setenv "MT_TEST_RUN_DIR" work-area)
	    (setenv "MT_TEST_NAME" test-name)
	    (setenv "MT_ITEM_INFO" (conc itemdat))
	    (setenv "MT_RUNNAME"   runname)
	    (setenv "MT_MEGATEST"  megatest)
	    (setenv "PATH" (conc (getenv "PATH") ":" mt-bindir-path))
	    
	    (if (not (setup-for-run))
		(begin
		  (debug:print 0 "Failed to setup, exiting") 
		  (exit 1)))
	    ;; now can find our db
	    (set! db (open-db))
	    (change-directory work-area) 
	    (let ((runconfigf (conc  *toppath* "/runconfigs.config")))
	      (if (file-exists? runconfigf)
		  (setup-env-defaults db runconfigf run-id)
		  (debug:print 0 "WARNING: You do not have a run config file: " runconfigf)))
            ;; environment overrides are done *before* the remaining critical envars.
            (alist->env-vars env-ovrd)
	    (set-megatest-env-vars db run-id)
	    (set-item-env-vars itemdat)
            (save-environment-as-files "megatest")
	    (test-set-meta-info db run-id test-name itemdat)
	    (test-set-status! db run-id test-name "REMOTEHOSTSTART" "n/a" itemdat (args:get-arg "-m"))
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
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
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
609
610
611
612
613
614
					 (test-update-meta-info db run-id test-name itemdat minutes)
					 (if kill-job? 
					     (begin
					       (mutex-lock! m)
					       (let* ((pid (vector-ref exit-info 0)))
						 (if (number? pid)
						     (begin
						       (print "WARNING: Request received to kill job (attempt # " kill-tries ")")
						       ;;(cond
						       ;;((>   kill-tries 0) ; 2)
						       (let ((processes (cmd-run->list (conc "pgrep -l -P " pid))))
							 (for-each 
							  (lambda (p)
							    (let* ((parts  (string-split p))
								   (p-id   (if (> (length parts) 0)
									       (string->number (car parts))
									       #f)))
							      (if p-id
								  (begin
								    (print "Killing " (cadr parts) "; kill -9  " p-id)
								    (system (conc "kill -9 " p-id))))))
							  (car processes))
							 (system (conc "kill -9 " pid))))
						     ;;(let* ((ppid (process-group-id pid))
						     ;;       (kcmd (conc "pkill -9 -g " ppid)))
						     ;;  ;; (process-signal pid signal/term)
						     ;;  ;; (process-signal pid signal/kill)
						     ;;  (print "Attempting to kill pid " pid " and children in process group " ppid " with command:\n    " kcmd)
						     ;;  (print "Children:")
						     ;;  (system (conc "pgrep -g -l " ppid))
						     ;;  (system kcmd)
						     ;;  (sleep 1) ;; give it a rest
						     ;;  (test-set-status! db run-id test-name "KILLED"  "FAIL"
						     ;;       	     itemdat (args:get-arg "-m"))
						     ;;  (sqlite3:finalize! db)
						     ;;  (exit 1)))))
						     (begin
						       (print "WARNING: Request received to kill job but problem with process, attempting to kill manager process")
						       (test-set-status! db run-id test-name "KILLED"  "FAIL"
									 itemdat (args:get-arg "-m"))
						       (sqlite3:finalize! db)
						       (exit 1))))
					       ;;     (thread-terminate! job-thread)))
					       (set! kill-tries (+ 1 kill-tries))
					       (mutex-unlock! m)))
					 ;; (handle-exceptions
					       ;;  exn
					       ;;  (begin
					       ;;    (print "ERROR: Problem killing process " (vector-ref exit-info 0))
					       ;;    (abort exn))
					       ;;  (let* ((pid   (vector-ref exit-info 0))
					       ;;         ;; (pgid  (process-group-id pid))
					       ;;         ;; (cmd  (conc "pkill -9 -P " pgid))
					       ;;         )
					       ;;    ;; (print "Running \"" cmd "\"")
					       ;;    ;; (system cmd)
					       ;;    (print "Running \"kill -9 " pid "\"")
					       ;;    (system (conc "kill -9 " pid))
					       ;;    ;; (process-signal (vector-ref exit-info 0) signal/kill)
					       ;;    ))))
					 (sqlite3:finalize! db)
					 (thread-sleep! (+ 8 (random 4))) ;; add some jitter to the call home time to spread out the db accesses
					 (loop (calc-minutes)))))))
		   (th1          (make-thread monitorjob))
		   (th2          (make-thread runit)))
	      (set! job-thread th2)
	      (thread-start! th1)
	      (thread-start! th2)
	      (thread-join! th2)
	      (mutex-lock! m)
	      (set! db (open-db))
	      (let* ((testinfo (db:get-test-info db run-id test-name (item-list->path itemdat))))
		(if (not (equal? (db:test-get-state testinfo) "COMPLETED"))
		    (begin
		      (print "Test NOT logged as COMPLETED, (state=" (db:test-get-state testinfo) "), updating result")
		      (test-set-status! db run-id test-name
					(if kill-job? "KILLED" "COMPLETED")
					(if (vector-ref exit-info 1) ;; look at the exit-status
					    (if (and (not kill-job?) 
						     (eq? (vector-ref exit-info 2) 0))
						"PASS"
						"FAIL")
					    "FAIL") itemdat (args:get-arg "-m")))))
	      (mutex-unlock! m)
	      ;; (exec-results (cmd-run->list fullrunscript)) ;;  (list ">" (conc test-name "-run.log"))))
	      ;; (success      exec-results)) ;; (eq? (cadr exec-results) 0)))
	      (print "Output from running " fullrunscript ", pid " (vector-ref exit-info 0) " in work area " 
		     work-area ":\n====\n exit code " (vector-ref exit-info 2) "\n" "====\n")
	      (sqlite3:finalize! db)
	      (if (not (vector-ref exit-info 1))
		  (exit 4)))))
      (set! *didsomething* #t)))

(if (args:get-arg "-step")
    (if (not (getenv "MT_CMDINFO"))
	(begin
	  (print "ERROR: MT_CMDINFO env var not set, -step must be called *inside* a megatest invoked environment!")
	  (exit 5))
	(let* ((step      (args:get-arg "-step"))
	       (cmdinfo   (read (open-input-string (base64:base64-decode (getenv "MT_CMDINFO")))))
	       (testpath  (assoc/default 'testpath  cmdinfo))
	       (test-name (assoc/default 'test-name cmdinfo))
	       (runscript (assoc/default 'runscript cmdinfo))
	       (db-host   (assoc/default 'db-host   cmdinfo))
	       (run-id    (assoc/default 'run-id    cmdinfo))
	       (itemdat   (assoc/default 'itemdat   cmdinfo))
	       (db        #f)
	       (state    (args:get-arg ":state"))
	       (status   (args:get-arg ":status")))
	  (change-directory testpath)
	  (if (not (setup-for-run))
	      (begin
		(print "Failed to setup, exiting")
		(exit 1)))
	  (set! db (open-db))
	  (if (and state status)
	      (teststep-set-status! db run-id test-name step state status itemdat (args:get-arg "-m"))
	      (begin
		(print "ERROR: You must specify :state and :status with every call to -step")
		(exit 6)))
	  (sqlite3:finalize! db)
	  (set! *didsomething* #t))))

(if (or (args:get-arg "-setlog")       ;; since setting up is so costly lets piggyback on -test-status
	(args:get-arg "-set-toplog")
	(args:get-arg "-test-status")
	(args:get-arg "-runstep"))
    (if (not (getenv "MT_CMDINFO"))
	(begin
	  (print "ERROR: MT_CMDINFO env var not set, commands -test-status, -runstep and -setlog must be called *inside* a megatest environment!")
	  (exit 5))
	(let* ((startingdir (current-directory))
	       (cmdinfo   (read (open-input-string (base64:base64-decode (getenv "MT_CMDINFO")))))
	       (testpath  (assoc/default 'testpath  cmdinfo))
	       (test-name (assoc/default 'test-name cmdinfo))
	       (runscript (assoc/default 'runscript cmdinfo))
	       (db-host   (assoc/default 'db-host   cmdinfo))
	       (run-id    (assoc/default 'run-id    cmdinfo))
	       (itemdat   (assoc/default 'itemdat   cmdinfo))
	       (db        #f)
	       (state     (args:get-arg ":state"))
	       (status    (args:get-arg ":status")))
	  (change-directory testpath)
	  (if (not (setup-for-run))
	      (begin
		(print "Failed to setup, exiting")
		(exit 1)))
	  (set! db (open-db))
	  (if (args:get-arg "-setlog")
	      (test-set-log! db run-id test-name itemdat (args:get-arg "-setlog")))
	  (if (args:get-arg "-set-toplog")
	      (test-set-toplog! db run-id test-name (args:get-arg "-set-toplog")))
	  (if (args:get-arg "-runstep")
	      (if (null? remargs)
		  (begin
		    (print "ERROR: nothing specified to run!")
		    (sqlite3:finalize! db)
		    (exit 6))
		  (let* ((stepname   (args:get-arg "-runstep"))
			 (logprofile (args:get-arg "-logpro"))
			 (logfile    (conc stepname ".log"))
			 (cmd        (if (null? remargs) #f (car remargs)))
			 (params     (if cmd (cdr remargs) '()))
			 (exitstat   #f)
			 (shell      (last (string-split (get-environment-variable "SHELL") "/")))
			 (redir      (case (string->symbol shell)
				       ((tcsh csh ksh)    ">&")
				       ((zsh bash sh ash) "2>&1 >")))
			 (fullcmd    (conc "(" (string-intersperse 
						(cons cmd params) " ")
					   ") " redir " " logfile)))
		    ;; mark the start of the test
		    (teststep-set-status! db run-id test-name stepname "start" "n/a" itemdat (args:get-arg "-m"))
		    ;; close the db
		    (sqlite3:finalize! db)
		    ;; run the test step
		    (print "INFO: Running \"" fullcmd "\"")
		    (change-directory startingdir)
		    (set! exitstat (system fullcmd)) ;; cmd params))
		    (set! *globalexitstatus* exitstat)
		    (change-directory testpath)
		    ;; re-open the db
		    (set! db (open-db)) 
		    ;; run logpro if applicable ;; (process-run "ls" (list "/foo" "2>&1" "blah.log"))
		    (if logprofile
			(let* ((htmllogfile (conc stepname ".html"))
			       (oldexitstat exitstat)
			       (cmd         (string-intersperse (list "logpro" logprofile htmllogfile "<" logfile ">" (conc stepname "_logpro.log")) " ")))
			  (print "INFO: running \"" cmd "\"")
			  (change-directory startingdir)
			  (set! exitstat (system cmd))
			  (set! *globalexitstatus* exitstat) ;; no necessary
			  (change-directory testpath)
			  (test-set-log! db run-id test-name itemdat htmllogfile)))
		    (teststep-set-status! db run-id test-name stepname "end" exitstat itemdat (args:get-arg "-m"))
		    (sqlite3:finalize! db)







|











|







|
|








|










|





|

|

















|











|









|















|





|










|















|









|




















|











|







424
425
426
427
428
429
430
431
432
433
434
435
436
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
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
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
					 (test-update-meta-info db run-id test-name itemdat minutes)
					 (if kill-job? 
					     (begin
					       (mutex-lock! m)
					       (let* ((pid (vector-ref exit-info 0)))
						 (if (number? pid)
						     (begin
						       (debug:print 0 "WARNING: Request received to kill job (attempt # " kill-tries ")")
						       ;;(cond
						       ;;((>   kill-tries 0) ; 2)
						       (let ((processes (cmd-run->list (conc "pgrep -l -P " pid))))
							 (for-each 
							  (lambda (p)
							    (let* ((parts  (string-split p))
								   (p-id   (if (> (length parts) 0)
									       (string->number (car parts))
									       #f)))
							      (if p-id
								  (begin
								    (debug:print 0 "Killing " (cadr parts) "; kill -9  " p-id)
								    (system (conc "kill -9 " p-id))))))
							  (car processes))
							 (system (conc "kill -9 " pid))))
						     ;;(let* ((ppid (process-group-id pid))
						     ;;       (kcmd (conc "pkill -9 -g " ppid)))
						     ;;  ;; (process-signal pid signal/term)
						     ;;  ;; (process-signal pid signal/kill)
						     ;;  (debug:print 0 "Attempting to kill pid " pid " and children in process group " ppid " with command:\n    " kcmd)
						     ;;  (debug:print 0 "Children:")
						     ;;  (system (conc "pgrep -g -l " ppid))
						     ;;  (system kcmd)
						     ;;  (sleep 1) ;; give it a rest
						     ;;  (test-set-status! db run-id test-name "KILLED"  "FAIL"
						     ;;       	     itemdat (args:get-arg "-m"))
						     ;;  (sqlite3:finalize! db)
						     ;;  (exit 1)))))
						     (begin
						       (debug:print 0 "WARNING: Request received to kill job but problem with process, attempting to kill manager process")
						       (test-set-status! db run-id test-name "KILLED"  "FAIL"
									 itemdat (args:get-arg "-m"))
						       (sqlite3:finalize! db)
						       (exit 1))))
					       ;;     (thread-terminate! job-thread)))
					       (set! kill-tries (+ 1 kill-tries))
					       (mutex-unlock! m)))
					 ;; (handle-exceptions
					       ;;  exn
					       ;;  (begin
					       ;;    (debug:print 0 "ERROR: Problem killing process " (vector-ref exit-info 0))
					       ;;    (abort exn))
					       ;;  (let* ((pid   (vector-ref exit-info 0))
					       ;;         ;; (pgid  (process-group-id pid))
					       ;;         ;; (cmd  (conc "pkill -9 -P " pgid))
					       ;;         )
					       ;;    ;; (debug:print 0 "Running \"" cmd "\"")
					       ;;    ;; (system cmd)
					       ;;    (debug:print 0 "Running \"kill -9 " pid "\"")
					       ;;    (system (conc "kill -9 " pid))
					       ;;    ;; (process-signal (vector-ref exit-info 0) signal/kill)
					       ;;    ))))
					 (sqlite3:finalize! db)
					 (thread-sleep! (+ 8 (random 4))) ;; add some jitter to the call home time to spread out the db accesses
					 (loop (calc-minutes)))))))
		   (th1          (make-thread monitorjob))
		   (th2          (make-thread runit)))
	      (set! job-thread th2)
	      (thread-start! th1)
	      (thread-start! th2)
	      (thread-join! th2)
	      (mutex-lock! m)
	      (set! db (open-db))
	      (let* ((testinfo (db:get-test-info db run-id test-name (item-list->path itemdat))))
		(if (not (equal? (db:test-get-state testinfo) "COMPLETED"))
		    (begin
		      (debug:print 2 "Test NOT logged as COMPLETED, (state=" (db:test-get-state testinfo) "), updating result")
		      (test-set-status! db run-id test-name
					(if kill-job? "KILLED" "COMPLETED")
					(if (vector-ref exit-info 1) ;; look at the exit-status
					    (if (and (not kill-job?) 
						     (eq? (vector-ref exit-info 2) 0))
						"PASS"
						"FAIL")
					    "FAIL") itemdat (args:get-arg "-m")))))
	      (mutex-unlock! m)
	      ;; (exec-results (cmd-run->list fullrunscript)) ;;  (list ">" (conc test-name "-run.log"))))
	      ;; (success      exec-results)) ;; (eq? (cadr exec-results) 0)))
	      (debug:print 2 "Output from running " fullrunscript ", pid " (vector-ref exit-info 0) " in work area " 
		     work-area ":\n====\n exit code " (vector-ref exit-info 2) "\n" "====\n")
	      (sqlite3:finalize! db)
	      (if (not (vector-ref exit-info 1))
		  (exit 4)))))
      (set! *didsomething* #t)))

(if (args:get-arg "-step")
    (if (not (getenv "MT_CMDINFO"))
	(begin
	  (debug:print 0 "ERROR: MT_CMDINFO env var not set, -step must be called *inside* a megatest invoked environment!")
	  (exit 5))
	(let* ((step      (args:get-arg "-step"))
	       (cmdinfo   (read (open-input-string (base64:base64-decode (getenv "MT_CMDINFO")))))
	       (testpath  (assoc/default 'testpath  cmdinfo))
	       (test-name (assoc/default 'test-name cmdinfo))
	       (runscript (assoc/default 'runscript cmdinfo))
	       (db-host   (assoc/default 'db-host   cmdinfo))
	       (run-id    (assoc/default 'run-id    cmdinfo))
	       (itemdat   (assoc/default 'itemdat   cmdinfo))
	       (db        #f)
	       (state    (args:get-arg ":state"))
	       (status   (args:get-arg ":status")))
	  (change-directory testpath)
	  (if (not (setup-for-run))
	      (begin
		(debug:print 0 "Failed to setup, exiting")
		(exit 1)))
	  (set! db (open-db))
	  (if (and state status)
	      (teststep-set-status! db run-id test-name step state status itemdat (args:get-arg "-m"))
	      (begin
		(debug:print 0 "ERROR: You must specify :state and :status with every call to -step")
		(exit 6)))
	  (sqlite3:finalize! db)
	  (set! *didsomething* #t))))

(if (or (args:get-arg "-setlog")       ;; since setting up is so costly lets piggyback on -test-status
	(args:get-arg "-set-toplog")
	(args:get-arg "-test-status")
	(args:get-arg "-runstep"))
    (if (not (getenv "MT_CMDINFO"))
	(begin
	  (debug:print 0 "ERROR: MT_CMDINFO env var not set, commands -test-status, -runstep and -setlog must be called *inside* a megatest environment!")
	  (exit 5))
	(let* ((startingdir (current-directory))
	       (cmdinfo   (read (open-input-string (base64:base64-decode (getenv "MT_CMDINFO")))))
	       (testpath  (assoc/default 'testpath  cmdinfo))
	       (test-name (assoc/default 'test-name cmdinfo))
	       (runscript (assoc/default 'runscript cmdinfo))
	       (db-host   (assoc/default 'db-host   cmdinfo))
	       (run-id    (assoc/default 'run-id    cmdinfo))
	       (itemdat   (assoc/default 'itemdat   cmdinfo))
	       (db        #f)
	       (state     (args:get-arg ":state"))
	       (status    (args:get-arg ":status")))
	  (change-directory testpath)
	  (if (not (setup-for-run))
	      (begin
		(debug:print 0 "Failed to setup, exiting")
		(exit 1)))
	  (set! db (open-db))
	  (if (args:get-arg "-setlog")
	      (test-set-log! db run-id test-name itemdat (args:get-arg "-setlog")))
	  (if (args:get-arg "-set-toplog")
	      (test-set-toplog! db run-id test-name (args:get-arg "-set-toplog")))
	  (if (args:get-arg "-runstep")
	      (if (null? remargs)
		  (begin
		    (debug:print 0 "ERROR: nothing specified to run!")
		    (sqlite3:finalize! db)
		    (exit 6))
		  (let* ((stepname   (args:get-arg "-runstep"))
			 (logprofile (args:get-arg "-logpro"))
			 (logfile    (conc stepname ".log"))
			 (cmd        (if (null? remargs) #f (car remargs)))
			 (params     (if cmd (cdr remargs) '()))
			 (exitstat   #f)
			 (shell      (last (string-split (get-environment-variable "SHELL") "/")))
			 (redir      (case (string->symbol shell)
				       ((tcsh csh ksh)    ">&")
				       ((zsh bash sh ash) "2>&1 >")))
			 (fullcmd    (conc "(" (string-intersperse 
						(cons cmd params) " ")
					   ") " redir " " logfile)))
		    ;; mark the start of the test
		    (teststep-set-status! db run-id test-name stepname "start" "n/a" itemdat (args:get-arg "-m"))
		    ;; close the db
		    (sqlite3:finalize! db)
		    ;; run the test step
		    (debug:print 2 "INFO: Running \"" fullcmd "\"")
		    (change-directory startingdir)
		    (set! exitstat (system fullcmd)) ;; cmd params))
		    (set! *globalexitstatus* exitstat)
		    (change-directory testpath)
		    ;; re-open the db
		    (set! db (open-db)) 
		    ;; run logpro if applicable ;; (process-run "ls" (list "/foo" "2>&1" "blah.log"))
		    (if logprofile
			(let* ((htmllogfile (conc stepname ".html"))
			       (oldexitstat exitstat)
			       (cmd         (string-intersperse (list "logpro" logprofile htmllogfile "<" logfile ">" (conc stepname "_logpro.log")) " ")))
			  (debug:print 2 "INFO: running \"" cmd "\"")
			  (change-directory startingdir)
			  (set! exitstat (system cmd))
			  (set! *globalexitstatus* exitstat) ;; no necessary
			  (change-directory testpath)
			  (test-set-log! db run-id test-name itemdat htmllogfile)))
		    (teststep-set-status! db run-id test-name stepname "end" exitstat itemdat (args:get-arg "-m"))
		    (sqlite3:finalize! db)
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
				((number? status)       (if (equal? status 0) "PASS" "FAIL"))
				((string->number status)(if (equal? (string->number status) 0) "PASS" "FAIL"))
				(else status))))
		(test-set-status! db run-id test-name state newstatus itemdat (args:get-arg "-m")))
	      (if (and state status)
		  (if (not (args:get-arg "-setlog"))
		      (begin
			(print "ERROR: You must specify :state and :status with every call to -test-status\n" help)
			(sqlite3:finalize! db)
			(exit 6)))))
	  (sqlite3:finalize! db)
	  (set! *didsomething* #t))))

(if (args:get-arg "-showkeys")
    (let ((db #f)
	  (keys #f))
      (if (not (setup-for-run))
	  (begin
	    (print "Failed to setup, exiting")
	    (exit 1)))
      (set! db (open-db))
      (set! keys (db-get-keys db))
      (print "Keys: " (string-intersperse (map key:get-fieldname keys) ", "))
      (sqlite3:finalize! db)
      (set! *didsomething* #t)))

(if (args:get-arg "-gui")
    (begin
      (print "Look at the dashboard for now")
      ;; (megatest-gui)
      (set! *didsomething* #t)))

(if (not *didsomething*)
    (print help))

(if (not (eq? *globalexitstatus* 0))
    (if (or (args:get-arg "-runtests")(args:get-arg "-runall"))
        (begin
           (print "NOTE: Subprocesses with non-zero exit code detected: " *globalexitstatus*)
           (exit 0))
        (case *globalexitstatus*
         ((0)(exit 0))
         ((1)(exit 1))
         ((2)(exit 2))
         (else (exit 3)))))







|










|



|





|




|




|






627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
				((number? status)       (if (equal? status 0) "PASS" "FAIL"))
				((string->number status)(if (equal? (string->number status) 0) "PASS" "FAIL"))
				(else status))))
		(test-set-status! db run-id test-name state newstatus itemdat (args:get-arg "-m")))
	      (if (and state status)
		  (if (not (args:get-arg "-setlog"))
		      (begin
			(debug:print 0 "ERROR: You must specify :state and :status with every call to -test-status\n" help)
			(sqlite3:finalize! db)
			(exit 6)))))
	  (sqlite3:finalize! db)
	  (set! *didsomething* #t))))

(if (args:get-arg "-showkeys")
    (let ((db #f)
	  (keys #f))
      (if (not (setup-for-run))
	  (begin
	    (debug:print 0 "Failed to setup, exiting")
	    (exit 1)))
      (set! db (open-db))
      (set! keys (db-get-keys db))
      (debug:print 1 "Keys: " (string-intersperse (map key:get-fieldname keys) ", "))
      (sqlite3:finalize! db)
      (set! *didsomething* #t)))

(if (args:get-arg "-gui")
    (begin
      (debug:print 0 "Look at the dashboard for now")
      ;; (megatest-gui)
      (set! *didsomething* #t)))

(if (not *didsomething*)
    (debug:print 0 help))

(if (not (eq? *globalexitstatus* 0))
    (if (or (args:get-arg "-runtests")(args:get-arg "-runall"))
        (begin
           (debug:print 0 "NOTE: Subprocesses with non-zero exit code detected: " *globalexitstatus*)
           (exit 0))
        (case *globalexitstatus*
         ((0)(exit 0))
         ((1)(exit 1))
         ((2)(exit 2))
         (else (exit 3)))))