Megatest

Check-in [3c2cad2a4d]
Login
Overview
Comment:fixe for end of run detection when last test is dead or dnr and remove runs without data
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | v1.65
Files: files | file ages | folders
SHA1: 3c2cad2a4da2f9d864bf42105e303298773d2df1
User & Date: pjhatwal on 2020-07-02 11:42:26
Other Links: branch diff | manifest | tags
Context
2020-07-02
16:25
added busy handler to mtutil process check-in: 376a95bd80 user: pjhatwal tags: v1.65, v1.6554
11:42
fixe for end of run detection when last test is dead or dnr and remove runs without data check-in: 3c2cad2a4d user: pjhatwal tags: v1.65
2020-06-16
12:10
bumped version, merged fork check-in: 5b46e5fdf9 user: mmgraham tags: v1.65, v1.6553
Changes

Modified db.scm from [17795efac6] to [ebea7e9d0d].

1745
1746
1747
1748
1749
1750
1751
1752

1753
1754
1755
1756
1757
1758
1759
                        (debug:print 0 *default-log-port* "INFO: test " test-id " actually passed, so marking PASS not DEAD")
                        (db:set-state-status-and-roll-up-items dbstruct run-id test-id 'foo "COMPLETED" "PASS" "Test stopped responding but it has PASSED; marking it PASS in the DB.")
                      )
                      (begin
                      (debug:print 0 *default-log-port* "INFO: test " test-id " final state/status is not COMPLETED/PASS. It is " result)
                      (db:set-state-status-and-roll-up-items dbstruct run-id test-id 'foo "COMPLETED" "DEAD" "Test stopped responding while in RUNNING or REMOTEHOSTSTART; presumed dead.")
                      )
                     )

                    )
                  )
                  all-ids)
             )
         )
       )
     )







|
>







1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
                        (debug:print 0 *default-log-port* "INFO: test " test-id " actually passed, so marking PASS not DEAD")
                        (db:set-state-status-and-roll-up-items dbstruct run-id test-id 'foo "COMPLETED" "PASS" "Test stopped responding but it has PASSED; marking it PASS in the DB.")
                      )
                      (begin
                      (debug:print 0 *default-log-port* "INFO: test " test-id " final state/status is not COMPLETED/PASS. It is " result)
                      (db:set-state-status-and-roll-up-items dbstruct run-id test-id 'foo "COMPLETED" "DEAD" "Test stopped responding while in RUNNING or REMOTEHOSTSTART; presumed dead.")
                      )
                     ) ;;call end of eud of run detection for posthook
                     (launch:end-of-run-check run-id)
                    )
                  )
                  all-ids)
             )
         )
       )
     )

Modified launch.scm from [b161d3487b] to [52a1e3ae7e].

907
908
909
910
911
912
913
914
915
916
917
918



919


920



921
922
923
924
925
926
927
(define (launch:end-of-run-check run-id )
    (let*	((not-completed-cnt (rmt:get-not-completed-cnt run-id))  
           (running-cnt (rmt:get-count-tests-running-for-run-id run-id))
           (all-test-launched (rmt:get-var (conc "lunch-complete-" run-id)))
           (current-state (rmt:get-run-state run-id))
           (current-status (rmt:get-run-status run-id)))
     ;;get-vars run-id to query metadata table to check if all completed. if all-test-launched = yes then only not-completed-cnt = 0 means everyting is completed if no entry found in the table do nothing 
     (debug:print 0 *default-log-port* "rollup run state/status")                      
     (rmt:set-state-status-and-roll-up-run  run-id current-state current-status)
     (runs:update-junit-test-reporter-xml run-id) 
     (cond 
       ((and all-test-launched (eq? not-completed-cnt 0) (equal? all-test-launched "yes" ))



           	(debug:print 0 *default-log-port* "look for  post hook.")


          	(runs:run-post-hook run-id))



        ((> running-cnt 3) 
        	  (debug:print 0 *default-log-port* "There are " running-cnt " tests running." ))
        ((> running-cnt 0)
            (debug:print 0 *default-log-port* "running cnt > 0 but <= 3 kill-running-tests-if-dead" )
   				  (let ((kill-cnt (launch:kill-tests-if-dead run-id)))
           			(if (and all-test-launched  (equal? all-test-launched "yes") (eq? kill-cnt running-cnt))
           					(launch:end-of-run-check run-id)))) ;;todo







|




>
>
>
|
>
>
|
>
>
>







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
(define (launch:end-of-run-check run-id )
    (let*	((not-completed-cnt (rmt:get-not-completed-cnt run-id))  
           (running-cnt (rmt:get-count-tests-running-for-run-id run-id))
           (all-test-launched (rmt:get-var (conc "lunch-complete-" run-id)))
           (current-state (rmt:get-run-state run-id))
           (current-status (rmt:get-run-status run-id)))
     ;;get-vars run-id to query metadata table to check if all completed. if all-test-launched = yes then only not-completed-cnt = 0 means everyting is completed if no entry found in the table do nothing 
     (debug:print 0 *default-log-port* "Running test cnt :" running-cnt)                      
     (rmt:set-state-status-and-roll-up-run  run-id current-state current-status)
     (runs:update-junit-test-reporter-xml run-id) 
     (cond 
       ((and all-test-launched (eq? not-completed-cnt 0) (equal? all-test-launched "yes" ))
                (if (and (equal? (rmt:get-var (conc "end-of-run-" run-id)) "no") (common:simple-lock (conc "endOfRun" run-id)))
                (begin
           	(debug:print 4 *default-log-port* "look for  post hook. currseconds: " (current-seconds) " EOR " (rmt:get-var (conc "end-of-run-" run-id)))
                (debug:print 0 *default-log-port* "End of Run Detected.")
                (rmt:set-var (conc "end-of-run-" run-id) "yes")
                ;(thread-sleep! 10)
          	(runs:run-post-hook run-id)
                (debug:print 4 *default-log-port* "currseconds: " (current-seconds)" eor: " (rmt:get-var (conc "end-of-run-" run-id)))
                (common:simple-unlock (conc "endOfRun" run-id)))
                 (debug:print 0 *default-log-port* "End of Run Detected but not running post hook. This should happen when eor is set to yes. This will happen only when 2 tests exit at smae time. eor= " (rmt:get-var (conc "end-of-run-" run-id)))))
        ((> running-cnt 3) 
        	  (debug:print 0 *default-log-port* "There are " running-cnt " tests running." ))
        ((> running-cnt 0)
            (debug:print 0 *default-log-port* "running cnt > 0 but <= 3 kill-running-tests-if-dead" )
   				  (let ((kill-cnt (launch:kill-tests-if-dead run-id)))
           			(if (and all-test-launched  (equal? all-test-launched "yes") (eq? kill-cnt running-cnt))
           					(launch:end-of-run-check run-id)))) ;;todo

Modified megatest-version.scm from [c33ef84182] to [0b51a1301e].

16
17
18
19
20
21
22
23
;;     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

;; Always use two or four digit decimal
;; 1.01, 1.02...1.10,1.11,1.1101 ... 1.99,2.00..

(declare (unit megatest-version))

(define megatest-version 1.6553)







|
16
17
18
19
20
21
22
23
;;     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

;; Always use two or four digit decimal
;; 1.01, 1.02...1.10,1.11,1.1101 ... 1.99,2.00..

(declare (unit megatest-version))

(define megatest-version 1.6554)

Modified runs.scm from [815542adba] to [8eb0eeffbc].

530
531
532
533
534
535
536









537
538
539
540
541
542
543
    (runs:update-all-test_meta #f)

    ;; run the run prehook if there are no tests yet run for this run:
    ;;
    (runs:run-pre-hook run-id)
    ;; mark all test launced flag as false in the meta table 
    (rmt:set-var (conc "lunch-complete-" run-id) "no")









    (rmt:set-run-state-status run-id "new" "n/a")
    ;; now add non-directly referenced dependencies (i.e. waiton)
    ;;======================================================================
    ;; refactoring this block into tests:get-full-data
    ;;
    ;; What happended, this code is now duplicated in tests!?
    ;;







>
>
>
>
>
>
>
>
>







530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
    (runs:update-all-test_meta #f)

    ;; run the run prehook if there are no tests yet run for this run:
    ;;
    (runs:run-pre-hook run-id)
    ;; mark all test launced flag as false in the meta table 
    (rmt:set-var (conc "lunch-complete-" run-id) "no")
    (debug:print-info 1 *default-log-port* "Setting end-of-run to no")
    (let* ((config-reruns      (let ((x (configf:lookup *configdat* "setup" "reruns")))
			       (if x (string->number x) #f)))
	  (config-rerun-cnt (if config-reruns
			config-reruns
			1)))
    (if (eq? config-rerun-cnt run-count)
      (rmt:set-var (conc "end-of-run-" run-id) "no")))

    (rmt:set-run-state-status run-id "new" "n/a")
    ;; now add non-directly referenced dependencies (i.e. waiton)
    ;;======================================================================
    ;; refactoring this block into tests:get-full-data
    ;;
    ;; What happended, this code is now duplicated in tests!?
    ;;
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
	    (if (> run-count 0) ;; handle reruns
		(begin
		  (if (not (hash-table-ref/default flags "-preclean" #f))
		      (hash-table-set! flags "-preclean" #t))
		  (if (not (hash-table-ref/default flags "-rerun" #f))
		      (hash-table-set! flags "-rerun" "STUCK/DEAD,n/a,ZERO_ITEMS"))
		  ;; recursive call to self
      (runs:run-tests target runname test-patts user flags run-count: (- run-count 1)))))

	  (debug:print-info 0 *default-log-port* "No tests to run")))
    (debug:print-info 4 *default-log-port* "All done by here")
    ;; TODO: try putting post hook call here
    ;(if (eq? run-count 0)
    ; (begin  
    ;  (debug:print-info 0 *default-log-port* "Calling Post Hook")  
    ;  (debug:print-info 2 *default-log-port* " run-count " run-count)
    ;  (runs:run-post-hook run-id))
    ;  (debug:print-info 2 *default-log-port* "Not calling post hook runcount = " run-count ))   
    (rmt:tasks-set-state-given-param-key task-key "done")

    ;; (sqlite3:finalize! tasks-db)
    ))


;; loop logic. These are used in runs:run-tests-queue to make it a bit more readable.
;;
;; If reg not full and have items in tal then loop with (car tal)(cdr tal) reg reruns







|
>



|
<
<




>







692
693
694
695
696
697
698
699
700
701
702
703
704


705
706
707
708
709
710
711
712
713
714
715
716
	    (if (> run-count 0) ;; handle reruns
		(begin
		  (if (not (hash-table-ref/default flags "-preclean" #f))
		      (hash-table-set! flags "-preclean" #t))
		  (if (not (hash-table-ref/default flags "-rerun" #f))
		      (hash-table-set! flags "-rerun" "STUCK/DEAD,n/a,ZERO_ITEMS"))
		  ;; recursive call to self
      (runs:run-tests target runname test-patts user flags run-count: (- run-count 1)))
                  (launch:end-of-run-check run-id)))
	  (debug:print-info 0 *default-log-port* "No tests to run")))
    (debug:print-info 4 *default-log-port* "All done by here")
    ;; TODO: try putting post hook call here
      


    ;  (debug:print-info 2 *default-log-port* " run-count " run-count)
    ;  (runs:run-post-hook run-id))
    ;  (debug:print-info 2 *default-log-port* "Not calling post hook runcount = " run-count ))   
    (rmt:tasks-set-state-given-param-key task-key "done")
     
    ;; (sqlite3:finalize! tasks-db)
    ))


;; loop logic. These are used in runs:run-tests-queue to make it a bit more readable.
;;
;; If reg not full and have items in tal then loop with (car tal)(cdr tal) reg reruns
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
	  (debug:print-info 0 *default-log-port* "Have leftovers!")
	  (loop (car reg)(cdr reg) '() reruns))
	 (else
          (debug:print-info 4 *default-log-port* "cond branch - "  "rtq-9")
	  (debug:print-info 4 *default-log-port* "Exiting loop with...\n  hed=" hed "\n  tal=" tal "\n  reruns=" reruns))
	 ))) ;; end loop on sorted test names
    ;; this is the point where everything is launched and now you can mark the run in metadata table as all launched 
    (rmt:set-var (conc "lunch-complete-" run-id) "yes")
  
    ;; now *if* -run-wait we wait for all tests to be done
    ;; Now wait for any RUNNING tests to complete (if in run-wait mode)
    (thread-sleep! 10) ;; I think there is a race condition here. Let states/statuses settle
    (let wait-loop ((num-running      (rmt:get-count-tests-running-for-run-id run-id))
		    (prev-num-running 0))
      ;; (debug:print-info 13 *default-log-port* "num-running=" num-running ", prev-num-running=" prev-num-running)
      (if (and (or (args:get-arg "-run-wait")







|
|







1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
	  (debug:print-info 0 *default-log-port* "Have leftovers!")
	  (loop (car reg)(cdr reg) '() reruns))
	 (else
          (debug:print-info 4 *default-log-port* "cond branch - "  "rtq-9")
	  (debug:print-info 4 *default-log-port* "Exiting loop with...\n  hed=" hed "\n  tal=" tal "\n  reruns=" reruns))
	 ))) ;; end loop on sorted test names
    ;; this is the point where everything is launched and now you can mark the run in metadata table as all launched 
    (rmt:set-var (conc "lunch-complete-" run-id) "yes")  
        
    ;; now *if* -run-wait we wait for all tests to be done
    ;; Now wait for any RUNNING tests to complete (if in run-wait mode)
    (thread-sleep! 10) ;; I think there is a race condition here. Let states/statuses settle
    (let wait-loop ((num-running      (rmt:get-count-tests-running-for-run-id run-id))
		    (prev-num-running 0))
      ;; (debug:print-info 13 *default-log-port* "num-running=" num-running ", prev-num-running=" prev-num-running)
      (if (and (or (args:get-arg "-run-wait")
2363
2364
2365
2366
2367
2368
2369






2370
2371
2372
2373
2374
2375
2376
                                          )
                                          (begin
                                            (debug:print 2 *default-log-port* "Not removing directory " rundir " because either it doesn't exist or has a bad name")
                                            (debug:print 2 *default-log-port* "Is /tmp/badname: " (string=  rundir "/tmp/badname"))
                                            (debug:print 2 *default-log-port* "Exists: " (file-exists? rundir))
                                            (debug:print 2 *default-log-port* "Has run-name: " (substring-index run-name rundir))
                                            (debug:print 2 *default-log-port* "Has target: " (substring-index target rundir))






                                          )
                                        )
                                      )

                                      (if (not (null? tal))
                                         (loop (car tal)(cdr tal)))))))
				(rmt:update-run-stats run-id (rmt:get-raw-run-stats run-id)))







>
>
>
>
>
>







2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
                                          )
                                          (begin
                                            (debug:print 2 *default-log-port* "Not removing directory " rundir " because either it doesn't exist or has a bad name")
                                            (debug:print 2 *default-log-port* "Is /tmp/badname: " (string=  rundir "/tmp/badname"))
                                            (debug:print 2 *default-log-port* "Exists: " (file-exists? rundir))
                                            (debug:print 2 *default-log-port* "Has run-name: " (substring-index run-name rundir))
                                            (debug:print 2 *default-log-port* "Has target: " (substring-index target rundir))
                                            ;;PJH remove record from db no need to cleanup directory
                                            (case mode
                                               ((remove-data-only)(mt:test-set-state-status-by-id (db:test-get-run_id test)(db:test-get-id test) (db:test-get-state test)(db:test-get-status test) #f))
                                               ((archive-remove)  (mt:test-set-state-status-by-id (db:test-get-run_id test)(db:test-get-id test) "ARCHIVED" #f #f))
                                               (else (rmt:delete-test-records (db:test-get-run_id test) (db:test-get-id test))))

                                          )
                                        )
                                      )

                                      (if (not (null? tal))
                                         (loop (car tal)(cdr tal)))))))
				(rmt:update-run-stats run-id (rmt:get-raw-run-stats run-id)))
2457
2458
2459
2460
2461
2462
2463


2464
2465
2466
2467
2468
2469
2470
2471
2472
                    (debug:print 1 *default-log-port* "Removing target " target "run: " run-name)
                    (if (not keep-records)
                      (begin
                        (debug:print 1 *default-log-port* "Removing DB records for the run.")
                        (rmt:delete-run run-id)
                        (rmt:delete-old-deleted-test-records))
                    )


	           (debug:print 1 *default-log-port* "Recursively removing links dir " linkspath)
                   (runs:recursive-delete-with-error-msg linkspath)

                   (for-each (lambda(runpath)
                       (debug:print 1 *default-log-port* "Recursively removing runs dir " runpath)
                       (runs:recursive-delete-with-error-msg runpath)
                     )
                     runpaths
                   )







>
>
|
|







2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
                    (debug:print 1 *default-log-port* "Removing target " target "run: " run-name)
                    (if (not keep-records)
                      (begin
                        (debug:print 1 *default-log-port* "Removing DB records for the run.")
                        (rmt:delete-run run-id)
                        (rmt:delete-old-deleted-test-records))
                    )
                    (if (not (equal?  linkspath "/does/not/exist/I"))
	               (begin 
                         (debug:print 1 *default-log-port* "Recursively removing links dir " linkspath)
                         (runs:recursive-delete-with-error-msg linkspath)))

                   (for-each (lambda(runpath)
                       (debug:print 1 *default-log-port* "Recursively removing runs dir " runpath)
                       (runs:recursive-delete-with-error-msg runpath)
                     )
                     runpaths
                   )
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
(define doc-template 
  '(*TOP*
    (*PI* xml "version='1.0'")
    (testsuite)))

(define (runs:update-junit-test-reporter-xml run-id)
  (let*	(
	 (junit-test-reporter	 (configf:lookup *configdat* "runs" "junit-test-reporter-xml"))
	 (junit-test-report-dir  (configf:lookup *configdat* "runs" "junit-test-report-dir"))
	 (xml-dir		 (if (and junit-test-reporter (equal? junit-test-reporter "yes" ))
				     (if junit-test-report-dir
					 junit-test-report-dir
					 (conc (getenv "MT_LINKTREE") "/" (getenv "MT_TARGET") "/" (getenv "MT_RUNNAME")))
				     #f))
	 (xml-ts-name		(if xml-dir
				    (conc (getenv "MT_TESTSUITENAME")"."(string-translate (getenv "MT_TARGET") "/" ".") "." (getenv "MT_RUNNAME"))
				    #f))
         (keyname               (if xml-ts-name
				    (common:get-signature xml-ts-name)
				    #f))
	 (xml-path		(if xml-dir
				    (conc xml-dir "/" keyname ".xml")
				    #f))

	 (test-data		(if xml-dir
				    (rmt:get-tests-for-run run-id "%" '() '() ;; run-id testpatt states statuses
							   #f #f ;; offset limit







|

|
|
|
|
|



|
<
<







2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767


2768
2769
2770
2771
2772
2773
2774
(define doc-template 
  '(*TOP*
    (*PI* xml "version='1.0'")
    (testsuite)))

(define (runs:update-junit-test-reporter-xml run-id)
  (let*	(
	 (junit-test-reporter	(configf:lookup *configdat* "runs" "junit-test-reporter-xml"))
	 (junit-test-report-dir  (configf:lookup *configdat* "runs" "junit-test-report-dir"))
	 (xml-dir		(if (and junit-test-reporter (equal? junit-test-reporter "yes" ))
				    (if junit-test-report-dir
					junit-test-report-dir
					(conc (getenv "MT_LINKTREE") "/" (getenv "MT_TARGET") "/" (getenv "MT_RUNNAME")))
				    #f))
	 (xml-ts-name		(if xml-dir
				    (conc (getenv "MT_TESTSUITENAME")"."(string-translate (getenv "MT_TARGET") "/" ".") "." (getenv "MT_RUNNAME"))
				    #f))
         (keyname               (if xml-ts-name (common:get-signature xml-ts-name) #f))


	 (xml-path		(if xml-dir
				    (conc xml-dir "/" keyname ".xml")
				    #f))

	 (test-data		(if xml-dir
				    (rmt:get-tests-for-run run-id "%" '() '() ;; run-id testpatt states statuses
							   #f #f ;; offset limit