Megatest

;; 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 1.6013)

	 (task-key           (conc (hash-table->alist flags) " " (get-host-name) " " (current-process-id)))
	 (tdbdat             (tasks:open-db)))

    (if (tasks:need-server run-id)(tasks:start-and-wait-for-server tdbdat run-id 10))

    (set-signal-handler! signal/int
			 (lambda (signum)
			   ;; (signal-mask! signum) ;; to mask or not? seems to cause issues in exiting
			   (print "Received signal " signum ", cleaning up before exit. Please wait...")
			   (let ((tdbdat (tasks:open-db)))
			     (rmt:tasks-set-state-given-param-key task-key "killed"))
			   (print "Killed by signal " signum ". Exiting")
			   (exit)))

    ;; register this run in monitor.db
    (rmt:tasks-add "run-tests" user target runname test-patts task-key) ;; params)
    (rmt:tasks-set-state-given-param-key task-key "running")
    (runs:set-megatest-env-vars run-id inkeys: keys inrunname: runname) ;; these may be needed by the launching process
    (if (file-exists? runconfigf)
	(setup-env-defaults runconfigf run-id *already-seen-runconfig-info* keyvals target)
	(debug:print 0 "WARNING: You do not have a run config file: " runconfigf))

    ;; Now generate all the tests lists
    (set! all-tests-registry (tests:get-all))
    (set! all-test-names     (hash-table-keys all-tests-registry))
    (set! test-names         (tests:filter-test-names all-test-names test-patts))

    ;; I think seeding required-tests with all test-names makes sense but lack analysis to back that up.
    ;;
    ;; (set! required-tests     (lset-intersection equal? (string-split test-patts ",") all-test-names))
    (set! required-tests     (lset-intersection equal? test-names all-test-names))
    
    ;; look up all tests matching the comma separated list of globs in
    ;; test-patts (using % as wildcard)

    ;; (set! test-names (delete-duplicates (tests:get-valid-tests *toppath* test-patts)))
    (debug:print-info 0 "tests search path: " (string-intersperse (tests:get-tests-search-path *configdat*) " "))
    (debug:print-info 0 "all tests:         " (string-intersperse (sort all-test-names string<) " "))
    (debug:print-info 0 "test names:        " (string-intersperse (sort test-names string<) " "))
    (debug:print-info 0 "required tests:    " (string-intersperse (sort required-tests string<) " "))

    ;; on the first pass or call to run-tests set FAILS to NOT_STARTED if
    ;; -keepgoing is specified
    (if (eq? *passnum* 0)
	(begin
	  ;; Is this still necessary? I think not. Unreachable tests are marked as such and 
	  ;; should not cause problems here.

             ;;    		   (not (member x allinqueue)))
             ;;    		 prereqstrs))
	     (give-up    #f))

	;; We can get here when a prereq has not been run due to *it* having a prereq that failed.
	;; We need to use this to dequeue this item as CANNOTRUN
	;; 
	(if (member 'toplevel testmode) ;; '(toplevel)) ;; NOTE: this probably should be (member 'toplevel testmode)
	    (for-each (lambda (prereq)
			(if (eq? (hash-table-ref/default test-registry prereq 'justfine) 'CANNOTRUN)
			    (set! give-up #t)))
		      prereqstrs))

	(if (and give-up
		 (not (and (null? tal)(null? reg))))

	 (num-running             (list-ref run-limits-info 1))
	 (num-running-in-jobgroup (list-ref run-limits-info 2)) 
	 (max-concurrent-jobs     (list-ref run-limits-info 3))
	 (job-group-limit         (list-ref run-limits-info 4))
	 (prereqs-not-met         (rmt:get-prereqs-not-met run-id waitons item-path testmode itemmap: itemmap))
	 ;; (prereqs-not-met         (mt:lazy-get-prereqs-not-met run-id waitons item-path mode: testmode itemmap: itemmap))
	 (fails                   (runs:calc-fails prereqs-not-met))
	 (non-completed           (filter (lambda (x)             ;; remove hed from not completed list, duh, of course it is not completed!
					    (not (equal? x hed)))
					  (runs:calc-not-completed prereqs-not-met)))
	 (loop-list               (list hed tal reg reruns))
	 ;; configure the load runner
	 (numcpus                 (common:get-num-cpus))
	 (maxload                 (string->number (or (configf:lookup *configdat* "jobtools" "maxload") "3")))
	 (waitdelay               (string->number (or (configf:lookup *configdat* "jobtools" "waitdelay") "60"))))
    (debug:print-info 4 "have-resources: " have-resources " prereqs-not-met: (" 
		      (string-intersperse 
		       (map (lambda (t)
			      (if (vector? t)
				  (conc (db:test-get-state t) "/" (db:test-get-status t))
				  (conc " WARNING: t is not a vector=" t )))
			    prereqs-not-met)
		       ", ") ") fails: " fails
		       "\nregistered? " (hash-table-ref/default test-registry (db:test-make-full-name test-name item-path) #f))
			    

    
    (if (and (not (null? prereqs-not-met))
	     (runs:lownoise (conc "waiting on tests " prereqs-not-met hed) 60))
	(debug:print-info 2 "waiting on tests; " (string-intersperse (runs:mixed-list-testname-and-testrec->list-of-strings prereqs-not-met) ", ")))

    ;; Don't know at this time if the test have been launched at some time in the past
    ;; i.e. is this a re-launch?

      ;; could have done hed tal here but doing car/cdr of newtal to rotate tests
      (list (car newtal)(cdr newtal) reg reruns))
     
     ;; This is the final stage, everything is in place so launch the test
     ;;
     ((and have-resources
	   (or (null? prereqs-not-met)
	       (and (member 'toplevel testmode) ;;  'toplevel)
		    (null? non-completed))))
      ;; (hash-table-delete! *max-tries-hash* (db:test-make-full-name test-name item-path))
      ;; we are going to reset all the counters for test retries by setting a new hash table
      ;; this means they will increment only when nothing can be run
      (set! *max-tries-hash* (make-hash-table))
      ;; well, first lets see if cpu load throttling is enabled. If so wait around until the
      ;; average cpu load is under the threshold before continuing

      ;; a message and drop hed from the items to be processed.
      ;; (runs:mixed-list-testname-and-testrec->list-of-strings prereqs-not-met)
      (if (and (not (null? prereqs-not-met))
	       (runs:lownoise (conc "waiting on tests " prereqs-not-met hed) 60))
	  (debug:print-info 1 "waiting on tests; " (string-intersperse 
						    (runs:mixed-list-testname-and-testrec->list-of-strings 
						     prereqs-not-met) ", ")))
      (if (or (null? fails)
	      (member 'toplevel testmode))
	  (begin
	    ;; couldn't run, take a breather
	    (if  (runs:lownoise "Waiting for more work to do..." 60)
		 (debug:print-info 0 "Waiting for more work to do..."))
	    (thread-sleep! 1)
	    (list (car newtal)(cdr newtal) reg reruns))
	  ;; the waiton is FAIL so no point in trying to run hed ever again

[ezsteps]
calcresults megatest -list-runs $MT_RUNNAME -target $MT_TARGET

[requirements]
waiton                      exit_0 exit_1  ez_exit2_fail  ez_fail        ez_pass              ezlog_fail \
       ezlog_fail_then_pass ezlog_pass     ezlog_warn     lineitem_fail  lineitem_pass        logpro_required_fail \
       manual_example       neverrun       priority_1     priority_10    priority_10_waiton_1 \
       priority_3           priority_4     priority_5     priority_6     priority_7           priority_8 \
       priority_9           runfirst       singletest     singletest2    sqlitespeed          test_mt_vars \
       ez_fail_quick        test1          test2

# This is a "toplevel" test, it does not require waitons to be non-FAIL to run
mode toplevel