Megatest

;;     GNU General Public License for more details.
;; 
;;     You should have received a copy of the GNU General Public License
;;     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

;;  strftime('%m/%d/%Y %H:%M:%S','now','localtime')

(use (prefix sqlite3 sqlite3:) srfi-1 posix regex regex-case srfi-69 (srfi 18) 
     posix-extras directory-utils pathname-expand typed-records format  sxml-serializer
     sxml-modifications matchable)

(declare (unit runs))
(declare (uses db))
(declare (uses common))
(declare (uses debugprint))
(declare (uses commonmod))
(declare (uses items))
(declare (uses runconfig))
(declare (uses tests))
(declare (uses server))
(declare (uses mt))
(declare (uses archive))
;; (declare (uses filedb))
(declare (uses mtargs))
(declare (uses rmtmod))
(declare (uses dbfile))

(use (prefix sqlite3 sqlite3:) srfi-1 posix regex regex-case srfi-69 (srfi 18) 
     posix-extras directory-utils pathname-expand typed-records format  sxml-serializer
     sxml-modifications matchable)



(include "common_records.scm")
(include "key_records.scm")
(include "db_records.scm")
(include "run_records.scm")
(include "test_records.scm")

;; (include "debugger.scm")

(import commonmod
	debugprint
	rmtmod
	dbfile
	(prefix mtargs args:))

;; use this struct to facilitate refactoring
;;

(defstruct runs:dat
  reglen regfull
  runname max-concurrent-jobs run-id
  test-patts required-tests test-registry

	(debug:print-info 1 *default-log-port* "Adding \"" (string-intersperse required-tests " ") "\" to the run queue"))
    ;; NOTE: these are all parent tests, items are not expanded yet.
    (debug:print-info 4 *default-log-port* "test-records=" (hash-table->alist test-records))
    (let ((reglen (configf:lookup *configdat* "setup" "runqueue")))
      (if (> (length (hash-table-keys test-records)) 0)
	  (let* ((keep-going        #t)
		 (run-queue-retries 5)
		;; (th1        (make-thread (lambda ()
		;; 			    (handle-exceptions
		;; 				exn
		;; 				(begin
		;; 				  (print-call-chain)
		;; 				  (print " message: " ((condition-property-accessor 'exn 'message) exn)))
		;; 			      (runs:run-tests-queue run-id runname test-records keyvals flags test-patts required-tests
		;; 						    (any->number reglen) all-tests-registry)))
		;; 			  "runs:run-tests-queue"))
		 (th2        (make-thread (lambda ()			 ;; BBQ: why are we visiting ALL runs here?	    
					    ;; (rmt:find-and-mark-incomplete-all-runs))))) CAN'T INTERRUPT IT ...
					    (let ((run-ids (rmt:get-all-run-ids)))
					      (for-each (lambda (run-id)
							  (if keep-going
							      (handle-exceptions
							       exn
							       (debug:print 0 *default-log-port* "error in calling find-and-mark-incomplete for run-id " run-id ", exn=" exn)
							       (rmt:find-and-mark-incomplete run-id #f)))) ;; ovr-deadtime))) ;; could be root of https://hsdes.intel.com/appstore/article/#/220546828/main -- Title: Megatest jobs show DEAD even though they are still running (1.64/27)
							run-ids)))
		 (run-ids (rmt:get-all-run-ids)))
	    #;(for-each (lambda (run-id)
			(if keep-going
			    (handle-exceptions
				exn
			      (debug:print 0 *default-log-port* "error in calling find-and-mark-incomplete for run-id " run-id ", exn=" exn)
			      (rmt:find-and-mark-incomplete run-id #f)))) ;; ovr-deadtime))) ;; could be root of https://hsdes.intel.com/appstore/article/#/220546828/main -- Title: Megatest jobs show DEAD even though they are still running (1.64/27)
		      run-ids)
					  "runs: mark-incompletes")))
	    ;; (thread-start! th1)
	    (thread-start! th2)
	    ;; (thread-join! th1)
	    ;; just do the main stuff in the main thread
	    (runs:run-tests-queue run-id runname test-records keyvals flags test-patts required-tests
								    (any->number reglen) all-tests-registry)
				  (any->number reglen) all-tests-registry)
	    (set! keep-going #f)
	    (thread-join! th2)
	    ;; if run-count > 0 call, set -preclean and -rerun STUCK/DEAD
	    (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" "ABORT,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)))
                (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 ))   

			 ", "))
      (thread-sleep! 0.051)
      (list hed tal reg reruns))
     
     ;; If no resources are available just kill time and loop again
     ;;
     ((not have-resources) ;; simply try again after waiting a second
      (if (runs:lownoise "no resources" 60)
      (if (runs:lownoise "no resources" 600)
	  (debug:print-info 1 *default-log-port* "no resources to run new tests, waiting ..."))

      ;; Have gone back and forth on this but db starvation is an issue.
      ;; wait one second before looking again to run jobs.
      (thread-sleep! 0.25)
      ;; (thread-sleep! 0.25)
      
      ;; new logic.
      ;; If it has been more than 10 seconds since we were last here don't wait at all
      ;; otherwise sleep 2 seconds to give db a rest and let dashboard read data
      (if (runs:lownoise "frequent-no-resources" 10)
	  (thread-sleep! 0.25) ;; no significant delay
	  (thread-sleep! 2))
      ;; 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)

         (registry-mutex        (make-mutex))
         (num-retries           0)
         (max-retries           (configf:lookup *configdat* "setup" "maxretries"))
         (max-concurrent-jobs   (configf:lookup-number *configdat* "setup" "max_concurrent_jobs" default: 50))
         (reglen                (if (number? reglen-in) reglen-in 1))
         (last-time-incomplete  (- (current-seconds) 900)) ;; force at least one clean up cycle
         (last-time-some-running (current-seconds))
	 (incoming-tests        '()) ;; queue up incoming tests here to tack on to tal when it gets low
         ;; (tdbdat                (tasks:open-db))
         (runsdat (make-runs:dat
                   ;; hed: hed
                   ;; tal: tal
                   ;; reg: reg
                   ;; reruns: reruns
                   reglen: reglen

		 (let* ((jobgroup              (runs:testdat-jobgroup testdat-in))
			(can-run-more-tests    (runs:dat-can-run-more-tests runsdat))
			(last-jobs-check-time  (runs:dat-last-jobs-check-time runsdat))
			(should-check-jobs     (match can-run-more-tests
						 ((can-run-more-flag num-running nr-in-jobgroup max-concurrent-jobs . params)
						  (if (< (- max-concurrent-jobs num-running) 25)
						      (begin
							(debug:print-info 0 *default-log-port*
							(debug:print-info 2 *default-log-port*
									  "less than 20 jobs headroom, ("max-concurrent-jobs
									  "-"num-running")>20. Forcing prelaunch check.")
							#t)
						      #f))
						 (else #f)))) ;; no record yet
		   (if should-check-jobs
		       (let loop-can-run-more

                                   (my-item-path (item-list->path my-itemdat))

                                   (newtestname (db:test-make-full-name hed my-item-path)))    ;; test names are unique on testname/item-path
                              (tests:testqueue-set-items!     new-test-record #f)
                              (tests:testqueue-set-itemdat!   new-test-record my-itemdat)
                              (tests:testqueue-set-item_path! new-test-record my-item-path)
                              (hash-table-set! test-records newtestname new-test-record)
			      ;; BUG: This next line sucks up a lot of horsepower
			      ;; (set! tal (append tal (list newtestname)))
			      ;; (set! tal (cons newtestname tal)) ;; 4/6/2023 - try using cons, does it matter if the test gets added at the beginning?
			      (set! incoming-tests (cons newtestname incoming-tests))
                              (set! tal (append tal (list newtestname)))))  ;; since these are itemized create new test names testname/itempath
			      ))  ;; since these are itemized create new test names testname/itempath
                          items-in-testpatt)))
          
          
	  (if (and (< (length tal) 20)
		   (not (null? incoming-tests)))
	      (begin
		(set! tal (append tal (reverse incoming-tests)))
		(set! incoming-tests '())))

	  ;; At this point we have possibly added items to tal but all must be handed off to 
	  ;; INNER COND logic. I think loop without rotating the queue 
	  ;; (loop hed tal reg reruns))
	  ;; (let ((newtal (append tal (list hed))))  ;; We should discard hed as it has been expanded into it's items? Yes, but only if this *is* an itemized test
	  ;; (loop (car newtal)(cdr newtal) reg reruns)
	  (if (null? tal)

	 (state-status (if (string? new-state-status) (string-split new-state-status ",") '(#f #f)))
	 (rp-mutex     (make-mutex))
	 (bup-mutex    (make-mutex))
         (keep-records (args:get-arg "-keep-records")) ;; used in conjunction with -remove-runs to keep the records, TODO: consolidate this with "mode".
	 (test-records '())) ;; for tasks that we wish to operate on all tests in one fell swoop

    (let* ((write-access-actions '(remove-runs set-state-status archive run-wait kill-runs))
           (dbfile             (conc  *toppath* "/.megatest/main.db"))
           (dbfile             (conc  *toppath* "/.mtdb/main.db"))
           (readonly-mode      (not (file-write-access? dbfile))))
      (when (and readonly-mode
                 (member action write-access-actions))
        (debug:print-error 0 *default-log-port* dbfile " is readonly.  Cannot proceed with action ["action"] in which write-access isrequired .")
        (exit 1)))
    
    (debug:print-info 4 *default-log-port* "runs:operate-on => Header: " header " action: " action " new-state-status: " new-state-status)