Megatest

Check-in [6f1893ddd7]
Login
Overview
Comment:Short circuit calculation of number tests running.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | v1.65
Files: files | file ages | folders
SHA1: 6f1893ddd776884caa933971d56b63f88334094b
User & Date: matt on 2020-08-17 06:20:15
Other Links: branch diff | manifest | tags
Context
2020-08-17
14:17
Replaced with-output-to-file with an explicit open/close due to posible file handles being left open. Fixed bug in runner time sharing. check-in: a1963bd5a9 user: mrwellan tags: v1.65, v1.6561
06:20
Short circuit calculation of number tests running. check-in: 6f1893ddd7 user: matt tags: v1.65
2020-08-16
22:07
Added glob caching for apparently expensive regexp execution check-in: b4a13d1106 user: matt tags: v1.65
Changes

Modified db.scm from [5d700139ad] to [c3242e998a].

3077
3078
3079
3080
3081
3082
3083
3084



3085
3086
3087
3088
3089

3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116



3117
3118
3119
3120
3121

3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
       (if newstatus  (sqlite3:execute db "UPDATE tests SET status=?  WHERE id=?;" newstatus  test-id))
       (if newcomment (sqlite3:execute db "UPDATE tests SET comment=? WHERE id=?;" newcomment ;; (sdb:qry 'getid newcomment)
				       test-id))))))
  (mt:process-triggers dbstruct run-id test-id newstate newstatus))

;; NEW BEHAVIOR: Count tests running in all runs!
;;
(define (db:get-count-tests-running dbstruct run-id)



  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (db)

     (sqlite3:first-result 
      db
      ;; WARNING BUG EDIT ME - merged from v1.55 - not sure what is right here ...
      ;; AND run_id NOT IN (SELECT id FROM runs WHERE state='deleted')
      "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '');"
      ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=?;" 
      ))))

;; NEW BEHAVIOR: Count tests running in only one run!
;;
(define (db:get-count-tests-actually-running dbstruct run-id)
  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (db)
     (sqlite3:first-result
      db
      ;; WARNING BUG EDIT ME - merged from v1.55 - not sure what is right here ...
      ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id NOT IN (SELECT id FROM runs WHERE state='deleted') AND NOT (uname = 'n/a' AND item_path = '');")
      "SELECT count(id) FROM tests WHERE state in ('RUNNING','REMOTEHOSTSTART','LAUNCHED') AND run_id=?;" 
      run-id)))) ;; NOT IN (SELECT id FROM runs WHERE state='deleted');")

;; NEW BEHAVIOR: Look only at single run with run-id
;; 
;; (define (db:get-running-stats dbstruct run-id)
(define (db:get-count-tests-running-for-run-id dbstruct run-id)



  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (db)

     (sqlite3:first-result
      db
      "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=?;" run-id))))

;; For a given testname how many items are running? Used to determine
;; probability for regenerating html
;; 
(define (db:get-count-tests-running-for-testname dbstruct run-id testname)
  (db:with-db
   dbstruct







|
>
>
>





>
|
<
<
<
<
<
<



















|
>
>
>
|
|
|
|
|
>
|
<
<







3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094






3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124


3125
3126
3127
3128
3129
3130
3131
       (if newstatus  (sqlite3:execute db "UPDATE tests SET status=?  WHERE id=?;" newstatus  test-id))
       (if newcomment (sqlite3:execute db "UPDATE tests SET comment=? WHERE id=?;" newcomment ;; (sdb:qry 'getid newcomment)
				       test-id))))))
  (mt:process-triggers dbstruct run-id test-id newstate newstatus))

;; NEW BEHAVIOR: Count tests running in all runs!
;;
(define (db:get-count-tests-running dbstruct run-id fastmode)
  (let* ((qry (if fastmode
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '') LIMIT 1;"
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '');")))
  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (db)
     (let* ((stmth (db:get-cache-stmth dbstruct  db qry)))
       (sqlite3:first-result stmth))))))







;; NEW BEHAVIOR: Count tests running in only one run!
;;
(define (db:get-count-tests-actually-running dbstruct run-id)
  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (db)
     (sqlite3:first-result
      db
      ;; WARNING BUG EDIT ME - merged from v1.55 - not sure what is right here ...
      ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id NOT IN (SELECT id FROM runs WHERE state='deleted') AND NOT (uname = 'n/a' AND item_path = '');")
      "SELECT count(id) FROM tests WHERE state in ('RUNNING','REMOTEHOSTSTART','LAUNCHED') AND run_id=?;" 
      run-id)))) ;; NOT IN (SELECT id FROM runs WHERE state='deleted');")

;; NEW BEHAVIOR: Look only at single run with run-id
;; 
;; (define (db:get-running-stats dbstruct run-id)
(define (db:get-count-tests-running-for-run-id dbstruct run-id fastmode)
  (let* ((qry (if fastmode
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=? LIMIT 1;"
		   "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=?;")))
    (db:with-db
     dbstruct
     run-id
     #f
     (lambda (db)
       (let* ((stmth (db:get-cache-stmth dbstruct db qry)))
	 (sqlite3:first-result stmth run-id))))))



;; For a given testname how many items are running? Used to determine
;; probability for regenerating html
;; 
(define (db:get-count-tests-running-for-testname dbstruct run-id testname)
  (db:with-db
   dbstruct

Modified launch.scm from [5fbad11dc6] to [56ecab23d3].

902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
;; new
;; 100% COMPLETED/ (PASS,FAIL,ABORT etc.) ==> COMPLETED / X where X is same as itemized rollup
;; > 3 RUNNING with not test_dead do nothing (run should already be RUNNING/ na
;; > 0 RUNNING and test_dead then send KILLREQ ==> COMPLETED
;; 0 RUNNING ==> this is actually the first condition, should not get here

(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) 







|
|







902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
;; new
;; 100% COMPLETED/ (PASS,FAIL,ABORT etc.) ==> COMPLETED / X where X is same as itemized rollup
;; > 3 RUNNING with not test_dead do nothing (run should already be RUNNING/ na
;; > 0 RUNNING and test_dead then send KILLREQ ==> COMPLETED
;; 0 RUNNING ==> this is actually the first condition, should not get here

(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 #f)) ;; fastmode=no
           (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) 

Modified rmt.scm from [db44f9a638] to [54042a29d0].

689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713

;; (define (rmt:get-run-ids-matching keynames target res)
;;   (rmt:send-receive #f 'get-run-ids-matching (list keynames target res)))

(define (rmt:get-prereqs-not-met run-id waitons ref-test-name ref-item-path #!key (mode '(normal))(itemmaps #f))
  (rmt:send-receive 'get-prereqs-not-met run-id (list run-id waitons ref-test-name ref-item-path mode itemmaps)))

(define (rmt:get-count-tests-running-for-run-id run-id)
  (rmt:send-receive 'get-count-tests-running-for-run-id run-id (list run-id)))

(define (rmt:get-not-completed-cnt run-id)
  (rmt:send-receive 'get-not-completed-cnt run-id (list run-id)))


;; Statistical queries

(define (rmt:get-count-tests-running run-id)
  (rmt:send-receive 'get-count-tests-running run-id (list run-id)))

(define (rmt:get-count-tests-running-for-testname run-id testname)
  (rmt:send-receive 'get-count-tests-running-for-testname run-id (list run-id testname)))

(define (rmt:get-count-tests-running-in-jobgroup run-id jobgroup)
  (rmt:send-receive 'get-count-tests-running-in-jobgroup run-id (list run-id jobgroup)))








|
|







|
|







689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713

;; (define (rmt:get-run-ids-matching keynames target res)
;;   (rmt:send-receive #f 'get-run-ids-matching (list keynames target res)))

(define (rmt:get-prereqs-not-met run-id waitons ref-test-name ref-item-path #!key (mode '(normal))(itemmaps #f))
  (rmt:send-receive 'get-prereqs-not-met run-id (list run-id waitons ref-test-name ref-item-path mode itemmaps)))

(define (rmt:get-count-tests-running-for-run-id run-id fastmode)
  (rmt:send-receive 'get-count-tests-running-for-run-id run-id (list run-id fastmode)))

(define (rmt:get-not-completed-cnt run-id)
  (rmt:send-receive 'get-not-completed-cnt run-id (list run-id)))


;; Statistical queries

(define (rmt:get-count-tests-running run-id fastmode)
  (rmt:send-receive 'get-count-tests-running run-id (list run-id fastmode)))

(define (rmt:get-count-tests-running-for-testname run-id testname)
  (rmt:send-receive 'get-count-tests-running-for-testname run-id (list run-id testname)))

(define (rmt:get-count-tests-running-in-jobgroup run-id jobgroup)
  (rmt:send-receive 'get-count-tests-running-in-jobgroup run-id (list run-id jobgroup)))

Modified runs.scm from [9bdec6c892] to [dbcca8ed98].

517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
        	  ((> (runs:dat-can-run-more-tests-count runsdat) 20) ;; original intent was - save cycles, wait a long time
		   (if (runs:lownoise "waiting on tasks" 60)(debug:print-info 2 *default-log-port* "waiting for tasks to complete, sleeping briefly ..."))
		   10)  ;; obviously haven't had any work to do for a while
		  (else
		   ;; if have a number for inter-test-delay, use it, else don't delay much, maybe even zero?
		   (configf:lookup-number *configdat* "setup" "inter-test-delay" default: 0.01))))
  
  (let* ((num-running             (rmt:get-count-tests-running run-id))
	 (num-running-in-jobgroup (rmt:get-count-tests-running-in-jobgroup run-id jobgroup))
	 (job-group-limit         (let ((jobg-count (configf:lookup *configdat* "jobgroups" jobgroup)))
				    (if (string? jobg-count)
					(string->number jobg-count)
					jobg-count))))
    (if (> (+ num-running num-running-in-jobgroup) 0)
	(runs:inc-can-run-more-tests-count runsdat)) ;; (set! *runs:can-run-more-tests-count* (+ *runs:can-run-more-tests-count* 1)))







|







517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
        	  ((> (runs:dat-can-run-more-tests-count runsdat) 20) ;; original intent was - save cycles, wait a long time
		   (if (runs:lownoise "waiting on tasks" 60)(debug:print-info 2 *default-log-port* "waiting for tasks to complete, sleeping briefly ..."))
		   10)  ;; obviously haven't had any work to do for a while
		  (else
		   ;; if have a number for inter-test-delay, use it, else don't delay much, maybe even zero?
		   (configf:lookup-number *configdat* "setup" "inter-test-delay" default: 0.01))))
  
  (let* ((num-running             (rmt:get-count-tests-running run-id #f)) ;; fastmode=no
	 (num-running-in-jobgroup (rmt:get-count-tests-running-in-jobgroup run-id jobgroup))
	 (job-group-limit         (let ((jobg-count (configf:lookup *configdat* "jobgroups" jobgroup)))
				    (if (string? jobg-count)
					(string->number jobg-count)
					jobg-count))))
    (if (> (+ num-running num-running-in-jobgroup) 0)
	(runs:inc-can-run-more-tests-count runsdat)) ;; (set! *runs:can-run-more-tests-count* (+ *runs:can-run-more-tests-count* 1)))
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
				       ))
				   extras)
				  extras)
				'())))
	     (waitons     (delete-duplicates (append (tests:testqueue-get-waitons test-record) extra-waits) equal?))
	     (newtal      (append tal (list hed)))
	     (regfull     (>= (length reg) reglen))
	     (num-running (rmt:get-count-tests-running-for-run-id run-id))
	     (testdat     (make-runs:testdat
			   hed: hed
			   tal: tal
			   reg: reg
			   reruns: reruns
			   test-record: test-record
			   test-name:   test-name







|







1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
				       ))
				   extras)
				  extras)
				'())))
	     (waitons     (delete-duplicates (append (tests:testqueue-get-waitons test-record) extra-waits) equal?))
	     (newtal      (append tal (list hed)))
	     (regfull     (>= (length reg) reglen))
	     (num-running (rmt:get-count-tests-running-for-run-id run-id #t)) ;; fastmode=yes
	     (testdat     (make-runs:testdat
			   hed: hed
			   tal: tal
			   reg: reg
			   reruns: reruns
			   test-record: test-record
			   test-name:   test-name
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980

1981
1982
1983
1984


1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
	 ))) ;; 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")
		   (equal? (configf:lookup *configdat* "setup" "run-wait") "yes"))
	       (> num-running 0))
	  (begin
	    ;; Here we mark any old defunct tests as incomplete. Do this every fifteen minutes
	    ;; (debug:print 0 *default-log-port* "Got here eh! num-running=" num-running " (> num-running 0) " (> num-running 0))
	    (if (> (current-seconds)(+ last-time-incomplete 900))
		(begin
		  (debug:print-info 0 *default-log-port* "Marking stuck tests as INCOMPLETE while waiting for run " run-id ". Running as pid " (current-process-id) " on " (get-host-name))

		  (set! last-time-incomplete (current-seconds))
		  (rmt:find-and-mark-incomplete run-id #f)))
	    (if (not (eq? num-running prev-num-running))
		(debug:print-info 0 *default-log-port* "run-wait specified, waiting on " num-running " tests in RUNNING, REMOTEHOSTSTART or LAUNCHED state at " (time->string (seconds->local-time (current-seconds)))))


	    (thread-sleep! 5)
	    ;; (wait-loop (rmt:get-count-tests-running-for-run-id run-id) num-running))))
	    (wait-loop (rmt:get-count-tests-running-for-run-id run-id) num-running))))
    ;; LET* ((test-record
    ;; we get here on "drop through". All done!
    ;; this is moved to runs:run-testes since this function is getting called twice to ensure everthing is completed. 
    ;; (debug:print-info 0 *default-log-port* "Calling Post Hook")    
    ;; (runs:run-post-hook run-id)
    (debug:print-info 1 *default-log-port* "All tests launched")))








|









|
|
>
|
|
<
|
>
>

|
|







1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983

1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
	 ))) ;; 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 #t)) ;; fastmode=yes
		    (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")
		   (equal? (configf:lookup *configdat* "setup" "run-wait") "yes"))
	       (> num-running 0))
	  (begin
	    ;; Here we mark any old defunct tests as incomplete. Do this every fifteen minutes
	    ;; (debug:print 0 *default-log-port* "Got here eh! num-running=" num-running " (> num-running 0) " (> num-running 0))
	    (if (> (current-seconds)(+ last-time-incomplete 900))
		(let ((actual-num-running (rmt:get-count-tests-running-for-run-id run-id #f))) ;; fastmode=no
		  (debug:print-info 0 *default-log-port* "Marking stuck tests as INCOMPLETE while waiting for run " run-id
				    ". Running as pid " (current-process-id) " on " (get-host-name))
		  (set! last-time-incomplete (current-seconds)) ;; FIXME, this might be causing slow down - use of set!
		  (rmt:find-and-mark-incomplete run-id #f)

		  (debug:print-info 0 *default-log-port* "run-wait specified, waiting on " actual-num-running
				    " tests in RUNNING, REMOTEHOSTSTART or LAUNCHED state at "
				    (time->string (seconds->local-time (current-seconds))))))
	    (thread-sleep! 5)
	    (wait-loop (rmt:get-count-tests-running-for-run-id run-id #t) ;; fastmode=yes
		       num-running))))
    ;; LET* ((test-record
    ;; we get here on "drop through". All done!
    ;; this is moved to runs:run-testes since this function is getting called twice to ensure everthing is completed. 
    ;; (debug:print-info 0 *default-log-port* "Calling Post Hook")    
    ;; (runs:run-post-hook run-id)
    (debug:print-info 1 *default-log-port* "All tests launched")))