Megatest

Diff
Login

Differences From Artifact [ad0152f814]:

To Artifact [d5c694d73c]:


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
	     (conc t)
	     (conc (db:test-get-testname t) ":" (db:test-get-state t) "/" (db:test-get-status t))))
       lst))

(define (runs:make-full-test-name testname itempath)
  (if (equal? itempath "") testname (conc testname "/" itempath)))

(define (runs:queue-next-hed tal reg n)
  (if (> (length reg) n)
      (car reg)

      (if (null? tal)
	  (car reg)
	  (car tal))))

(define (runs:queue-next-tal tal reg n)
  (if (> (length reg) n)
      ;; rotate
      (let ((nexttal (if (null? tal)
			 reg
			 (cdr tal))))
	(if (null? nexttal)
	    reg
	    nexttal))
      (if (null? reg)
	  '()
	  (cdr reg))))

(define (runs:queue-next-reg tal reg n)
  (if (> (length reg) n)
      (cdr reg)


      reg))
      

;; test-records is a hash table testname:item_path => vector < testname testconfig waitons priority items-info ... >
(define (runs:run-tests-queue run-id runname test-records keyvallst flags test-patts)
    ;; At this point the list of parent tests is expanded 
    ;; NB// Should expand items here and then insert into the run queue.
  (debug:print 5 "test-records: " test-records ", keyvallst: " keyvallst " flags: " (hash-table->alist flags))
  (let ((sorted-test-names (tests:sort-by-priority-and-waiton test-records))
	(test-registry    (make-hash-table))
	(registry-mutex   (make-mutex))
	(num-retries        0)
	(max-retries       (config-lookup *configdat* "setup" "maxretries"))
	(max-concurrent-jobs     (let ((mcj (config-lookup *configdat* "setup"     "max_concurrent_jobs")))
				    (if (and mcj (string->number mcj))
					(string->number mcj)
					1)))
	(reglen             10)) ;; length of the register queue ahead
    (set! max-retries (if (and max-retries (string->number max-retries))(string->number max-retries) 100))
    (if (not (null? sorted-test-names))
	(let loop ((hed         (car sorted-test-names))
		   (tal         (cdr sorted-test-names))
		   (reg         '()) ;; registered, put these at the head of tal 
		   (reruns      '()))
	  (if (not (null? reruns))(debug:print-info 4 "reruns=" reruns))
	  ;; (print "Top of loop, hed=" hed ", tal=" tal " ,reruns=" reruns)
	  (let* ((test-record (hash-table-ref test-records hed))
		 (test-name   (tests:testqueue-get-testname test-record))
		 (tconfig     (tests:testqueue-get-testconfig test-record))
		 (testmode    (let ((m (config-lookup tconfig "requirements" "mode")))
				(if m (string->symbol m) 'normal)))
		 (waitons     (tests:testqueue-get-waitons    test-record))
		 (priority    (tests:testqueue-get-priority   test-record))
		 (itemdat     (tests:testqueue-get-itemdat    test-record)) ;; itemdat can be a string, list or #f
		 (items       (tests:testqueue-get-items      test-record))
		 (item-path   (item-list->path itemdat))
		 (newtal      (append tal (list hed))))

	    ;; (if (> (length reg) 10)
	    ;;     (begin
	    ;;       (set! tal (cons hed tal))
	    ;;       (set! hed (car reg))
	    ;;       (set! reg (cdr reg))
	    ;;       (set! newtal tal)))
	    (debug:print 6







|
<
|
>
|
|
|

|
|
|
<
<
|
|

|
<
<
|

|
|

>
>
|
<















|


















|
>







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
	     (conc t)
	     (conc (db:test-get-testname t) ":" (db:test-get-state t) "/" (db:test-get-status t))))
       lst))

(define (runs:make-full-test-name testname itempath)
  (if (equal? itempath "") testname (conc testname "/" itempath)))

(define (runs:queue-next-hed tal reg n regful)

  (if regful
      (if (null? reg) ;; doesn't make sense, this is probably NOT the problem of the car
	  (car tal)
	  (car reg))
      (car tal)))

(define (runs:queue-next-tal tal reg n regful)
  (if regful
      tal


      (let ((newtal (cdr tal)))
	(if (null? newtal)
	    reg
	    newtal


	    ))))

(define (runs:queue-next-reg tal reg n regful)
  (if regful
      (cdr reg)
      (if (eq? (length tal) 1)
	  '()
	  reg)))


;; test-records is a hash table testname:item_path => vector < testname testconfig waitons priority items-info ... >
(define (runs:run-tests-queue run-id runname test-records keyvallst flags test-patts)
    ;; At this point the list of parent tests is expanded 
    ;; NB// Should expand items here and then insert into the run queue.
  (debug:print 5 "test-records: " test-records ", keyvallst: " keyvallst " flags: " (hash-table->alist flags))
  (let ((sorted-test-names (tests:sort-by-priority-and-waiton test-records))
	(test-registry    (make-hash-table))
	(registry-mutex   (make-mutex))
	(num-retries        0)
	(max-retries       (config-lookup *configdat* "setup" "maxretries"))
	(max-concurrent-jobs     (let ((mcj (config-lookup *configdat* "setup"     "max_concurrent_jobs")))
				    (if (and mcj (string->number mcj))
					(string->number mcj)
					1)))
	(reglen             2)) ;; length of the register queue ahead
    (set! max-retries (if (and max-retries (string->number max-retries))(string->number max-retries) 100))
    (if (not (null? sorted-test-names))
	(let loop ((hed         (car sorted-test-names))
		   (tal         (cdr sorted-test-names))
		   (reg         '()) ;; registered, put these at the head of tal 
		   (reruns      '()))
	  (if (not (null? reruns))(debug:print-info 4 "reruns=" reruns))
	  ;; (print "Top of loop, hed=" hed ", tal=" tal " ,reruns=" reruns)
	  (let* ((test-record (hash-table-ref test-records hed))
		 (test-name   (tests:testqueue-get-testname test-record))
		 (tconfig     (tests:testqueue-get-testconfig test-record))
		 (testmode    (let ((m (config-lookup tconfig "requirements" "mode")))
				(if m (string->symbol m) 'normal)))
		 (waitons     (tests:testqueue-get-waitons    test-record))
		 (priority    (tests:testqueue-get-priority   test-record))
		 (itemdat     (tests:testqueue-get-itemdat    test-record)) ;; itemdat can be a string, list or #f
		 (items       (tests:testqueue-get-items      test-record))
		 (item-path   (item-list->path itemdat))
		 (newtal      (append tal (list hed)))
		 (regfull     (> (length reg) reglen)))
	    ;; (if (> (length reg) 10)
	    ;;     (begin
	    ;;       (set! tal (cons hed tal))
	    ;;       (set! hed (car reg))
	    ;;       (set! reg (cdr reg))
	    ;;       (set! newtal tal)))
	    (debug:print 6
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
		 ;; Check item path against item-patts
		 ((not (tests:match test-patts (tests:testqueue-get-testname test-record) item-path)) ;; This test/itempath is not to be run
		  ;; else the run is stuck, temporarily or permanently
		  ;; but should check if it is due to lack of resources vs. prerequisites
		  (debug:print-info 1 "Skipping " (tests:testqueue-get-testname test-record) " " item-path " as it doesn't match " test-patts)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (runs:queue-next-hed tal reg reglen)
			    (runs:queue-next-tal tal reg reglen)
			    (runs:queue-next-reg tal reg reglen)
			    reruns)))
		 ;; Registry has been started for this test but has not yet completed
		 ;; this should be rare, the case where there are only a couple of tests and the db is slow
		 ;; delay a short while and continue
		 ;; ((eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f) 'start)
		 ;;  (thread-sleep! 0.01)
		 ;;  (loop (car newtal)(cdr newtal) reruns))







|
|
|







432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
		 ;; Check item path against item-patts
		 ((not (tests:match test-patts (tests:testqueue-get-testname test-record) item-path)) ;; This test/itempath is not to be run
		  ;; else the run is stuck, temporarily or permanently
		  ;; but should check if it is due to lack of resources vs. prerequisites
		  (debug:print-info 1 "Skipping " (tests:testqueue-get-testname test-record) " " item-path " as it doesn't match " test-patts)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (runs:queue-next-hed tal reg reglen regfull)
			    (runs:queue-next-tal tal reg reglen regfull)
			    (runs:queue-next-reg tal reg reglen regfull)
			    reruns)))
		 ;; Registry has been started for this test but has not yet completed
		 ;; this should be rare, the case where there are only a couple of tests and the db is slow
		 ;; delay a short while and continue
		 ;; ((eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f) 'start)
		 ;;  (thread-sleep! 0.01)
		 ;;  (loop (car newtal)(cdr newtal) reruns))
461
462
463
464
465
466
467
468
469
470



471
472
473
474
475
476
477
					   (cdb:tests-register-test *runremote* run-id test-name item-path)
		        		   (mutex-lock! registry-mutex)
					   (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'done)
		        		   (mutex-unlock! registry-mutex))
		        		 (conc test-name "/" item-path))))
		    (thread-start! th))
		  (runs:shrink-can-run-more-tests-count)   ;; DELAY TWEAKER (still needed?)
		  (loop (car newtal)
			(runs:queue-next-tal tal reg reglen)
			(append reg (list hed))



			reruns))
		 ;; At this point hed test registration must be completed.
		 ((eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f)
		       'start)
		  (debug:print-info 0 "Waiting on test registration(s): " (string-intersperse 
									   (filter (lambda (x)
										     (eq? (hash-table-ref/default test-registry x #f) 'start))







|
|
|
>
>
>







459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
					   (cdb:tests-register-test *runremote* run-id test-name item-path)
		        		   (mutex-lock! registry-mutex)
					   (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'done)
		        		   (mutex-unlock! registry-mutex))
		        		 (conc test-name "/" item-path))))
		    (thread-start! th))
		  (runs:shrink-can-run-more-tests-count)   ;; DELAY TWEAKER (still needed?)
		  (loop (runs:queue-next-hed tal reg reglen regfull)
			(runs:queue-next-tal tal reg reglen regfull)
			(let ((newl (append reg (list hed))))
			  (if regfull 
			      (cdr newl)
			      newl))
			reruns))
		 ;; At this point hed test registration must be completed.
		 ((eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f)
		       'start)
		  (debug:print-info 0 "Waiting on test registration(s): " (string-intersperse 
									   (filter (lambda (x)
										     (eq? (hash-table-ref/default test-registry x #f) 'start))
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
			   (and (eq? testmode 'toplevel)
				(null? non-completed))))
		  (run:test run-id runname keyvallst test-record flags #f)
		  (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'running)
		  (runs:shrink-can-run-more-tests-count)  ;; DELAY TWEAKER (still needed?)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (runs:queue-next-hed tal reg reglen)
			    (runs:queue-next-tal tal reg reglen)
			    (runs:queue-next-reg tal reg reglen)
			    reruns)))
		 (else ;; must be we have unmet prerequisites
		    (debug:print 4 "FAILS: " fails)
		    ;; If one or more of the prereqs-not-met are FAIL then we can issue
		    ;; a message and drop hed from the items to be processed.
		    (if (null? fails)
			(begin







|
|
|







492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
			   (and (eq? testmode 'toplevel)
				(null? non-completed))))
		  (run:test run-id runname keyvallst test-record flags #f)
		  (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'running)
		  (runs:shrink-can-run-more-tests-count)  ;; DELAY TWEAKER (still needed?)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (runs:queue-next-hed tal reg reglen regfull)
			    (runs:queue-next-tal tal reg reglen regfull)
			    (runs:queue-next-reg tal reg reglen regfull)
			    reruns)))
		 (else ;; must be we have unmet prerequisites
		    (debug:print 4 "FAILS: " fails)
		    ;; If one or more of the prereqs-not-met are FAIL then we can issue
		    ;; a message and drop hed from the items to be processed.
		    (if (null? fails)
			(begin
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
			    (if (vector? hed)
				(begin 
				  (debug:print 1 "WARN: Dropping test " (db:test-get-testname hed) "/" (db:test-get-item-path hed)
					       " from the launch list as it has prerequistes that are FAIL")
				  (runs:shrink-can-run-more-tests-count) ;; DELAY TWEAKER (still needed?)
				  ;; (thread-sleep! *global-delta*)
				  (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'removed)
				  (loop (runs:queue-next-hed tal reg reglen)
					(runs:queue-next-tal tal reg reglen)
					(runs:queue-next-reg tal reg reglen)
					(cons hed reruns)))
				(begin
				  (debug:print 1 "WARN: Test not processed correctly. Could be a race condition in your test implementation? " hed) ;;  " as it has prerequistes that are FAIL. (NOTE: hed is not a vector)")
				  (runs:shrink-can-run-more-tests-count) ;; DELAY TWEAKER (still needed?)
				  ;; (thread-sleep! (+ 0.01 *global-delta*))
				  (loop hed tal reg reruns))))))))) ;; END OF INNER COND
	     







|
|
|







516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
			    (if (vector? hed)
				(begin 
				  (debug:print 1 "WARN: Dropping test " (db:test-get-testname hed) "/" (db:test-get-item-path hed)
					       " from the launch list as it has prerequistes that are FAIL")
				  (runs:shrink-can-run-more-tests-count) ;; DELAY TWEAKER (still needed?)
				  ;; (thread-sleep! *global-delta*)
				  (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'removed)
				  (loop (runs:queue-next-hed tal reg reglen regfull)
					(runs:queue-next-tal tal reg reglen regfull)
					(runs:queue-next-reg tal reg reglen regfull)
					(cons hed reruns)))
				(begin
				  (debug:print 1 "WARN: Test not processed correctly. Could be a race condition in your test implementation? " hed) ;;  " as it has prerequistes that are FAIL. (NOTE: hed is not a vector)")
				  (runs:shrink-can-run-more-tests-count) ;; DELAY TWEAKER (still needed?)
				  ;; (thread-sleep! (+ 0.01 *global-delta*))
				  (loop hed tal reg reruns))))))))) ;; END OF INNER COND
	     
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
			 (hash-table-set! test-records newtestname new-test-record)
			 (set! tal (cons newtestname tal)))))) ;; since these are itemized create new test names testname/itempath
	       items)
	      (if (not (null? tal))
		  (begin
		    (debug:print-info 4 "End of items list, looping with next after short delay")
                    ;; (thread-sleep! (+ 0.01 *global-delta*))
		    (loop (runs:queue-next-hed tal reg reglen)
			  (runs:queue-next-tal tal reg reglen)
			  (runs:queue-next-reg tal reg reglen)
			  reruns))))

	     ;; if items is a proc then need to run items:get-items-from-config, get the list and loop 
	     ;;    - but only do that if resources exist to kick off the job
	     ((or (procedure? items)(eq? items 'have-procedure))
	      (let ((can-run-more    (runs:can-run-more-tests test-record max-concurrent-jobs)))
		(if (and (list? can-run-more)







|
|
|







551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
			 (hash-table-set! test-records newtestname new-test-record)
			 (set! tal (cons newtestname tal)))))) ;; since these are itemized create new test names testname/itempath
	       items)
	      (if (not (null? tal))
		  (begin
		    (debug:print-info 4 "End of items list, looping with next after short delay")
                    ;; (thread-sleep! (+ 0.01 *global-delta*))
		    (loop (runs:queue-next-hed tal reg reglen regfull)
			  (runs:queue-next-tal tal reg reglen regfull)
			  (runs:queue-next-reg tal reg reglen regfull)
			  reruns))))

	     ;; if items is a proc then need to run items:get-items-from-config, get the list and loop 
	     ;;    - but only do that if resources exist to kick off the job
	     ((or (procedure? items)(eq? items 'have-procedure))
	      (let ((can-run-more    (runs:can-run-more-tests test-record max-concurrent-jobs)))
		(if (and (list? can-run-more)
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
			    (begin
			      ;; TRY (if (> num-retries 100) ;; first 100 retries are low time cost
			      ;; TRY     (thread-sleep! (+ 2 *global-delta*))
			      ;; TRY     (thread-sleep! (+ 0.01 *global-delta*)))
			      (set! num-retries (+ num-retries 1))))
			(if (> num-retries  max-retries)
			    (if (not (null? tal))
				(loop (runs:queue-next-hed tal reg reglen)
				      (runs:queue-next-tal tal reg reglen)
				      (runs:queue-next-reg tal reg reglen)
				      reruns))
			    (loop (car newtal)(cdr newtal) reg reruns))) ;; an issue with prereqs not yet met?
		       ((and (not (null? fails))(eq? testmode 'normal))
			(debug:print-info 1 "test "  hed " (mode=" testmode ") has failed prerequisite(s); "
				     (string-intersperse (map (lambda (t)(conc (db:test-get-testname t) ":" (db:test-get-state t)"/"(db:test-get-status t))) fails) ", ")
				     ", removing it from to-do list")
			(if (not (null? tal))
			    (begin
                              ;; (thread-sleep! *global-delta*)
			      (loop (runs:queue-next-hed tal reg reglen)
				    (runs:queue-next-tal tal reg reglen)
				    (runs:queue-next-reg tal reg reglen)
				    (cons hed reruns)))))
		       (else
			(debug:print 8 "ERROR: No handler for this condition.")
			;; TRY (thread-sleep! (+ 1 *global-delta*))
			(loop (car newtal)(cdr newtal) reg reruns)))) ;; END OF IF CAN RUN MORE

		    ;; if can't run more just loop with next possible test







|
|
|









|
|
|







607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
			    (begin
			      ;; TRY (if (> num-retries 100) ;; first 100 retries are low time cost
			      ;; TRY     (thread-sleep! (+ 2 *global-delta*))
			      ;; TRY     (thread-sleep! (+ 0.01 *global-delta*)))
			      (set! num-retries (+ num-retries 1))))
			(if (> num-retries  max-retries)
			    (if (not (null? tal))
				(loop (runs:queue-next-hed tal reg reglen regfull)
				      (runs:queue-next-tal tal reg reglen regfull)
				      (runs:queue-next-reg tal reg reglen regfull)
				      reruns))
			    (loop (car newtal)(cdr newtal) reg reruns))) ;; an issue with prereqs not yet met?
		       ((and (not (null? fails))(eq? testmode 'normal))
			(debug:print-info 1 "test "  hed " (mode=" testmode ") has failed prerequisite(s); "
				     (string-intersperse (map (lambda (t)(conc (db:test-get-testname t) ":" (db:test-get-state t)"/"(db:test-get-status t))) fails) ", ")
				     ", removing it from to-do list")
			(if (not (null? tal))
			    (begin
                              ;; (thread-sleep! *global-delta*)
			      (loop (runs:queue-next-hed tal reg reglen regfull)
				    (runs:queue-next-tal tal reg reglen regfull)
				    (runs:queue-next-reg tal reg reglen regfull)
				    (cons hed reruns)))))
		       (else
			(debug:print 8 "ERROR: No handler for this condition.")
			;; TRY (thread-sleep! (+ 1 *global-delta*))
			(loop (car newtal)(cdr newtal) reg reruns)))) ;; END OF IF CAN RUN MORE

		    ;; if can't run more just loop with next possible test