Megatest

	(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))) ;; length of the register queue ahead
	(reglen                (if (number? reglen-in) reglen-in 1))
	(last-time-incomplete  (current-seconds))
	(last-time-incomplete  (- (current-seconds) 900)) ;; force at least one clean up cycle
	(last-time-some-running (current-seconds)))

    ;; Initialize the test-registery hash with tests that already have a record
    ;; convert state to symbol and use that as the hash value
    (for-each (lambda (trec)
		(let ((id (db:test-get-id        trec))
		      (tn (db:test-get-testname  trec))
		      (ip (db:test-get-item-path trec))
		      (st (db:test-get-state     trec)))
		  (if (not (equal? st "DELETED"))
		      (hash-table-set! test-registry (runs:make-full-test-name tn ip) (string->symbol st)))))
	      tests-info)
    (set! max-retries (if (and max-retries (string->number max-retries))(string->number max-retries) 100))

    (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))

      ;; Here we mark any old defunct tests as incomplete. Do this every fifteen minutes
      ;; (if (> (current-seconds)(+ last-time-incomplete 900))
      ;;     (begin
      ;;       (set! last-time-incomplete (current-seconds))
      ;;       (cdb:remote-run db:find-and-mark-incomplete #f)))
      (if (> (current-seconds)(+ last-time-incomplete 900))
          (begin
            (set! last-time-incomplete (current-seconds))
            (cdb:remote-run db:find-and-mark-incomplete #f)))

      ;; (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))
	     (jobgroup    (config-lookup tconfig "test_meta" "jobgroup"))
	     (testmode    (let ((m (config-lookup tconfig "requirements" "mode")))

;; fields are passing in through 
;; action:
;;    'remove-runs
;;    'set-state-status
;;
;; NB// should pass in keys?
;;
(define (runs:operate-on action target runnamepatt testpatt #!key (state #f)(status #f)(new-state-status #f))
(define (runs:operate-on action target runnamepatt testpatt #!key (state #f)(status #f)(new-state-status #f)(remove-data-only #f))
  (common:clear-caches) ;; clear all caches
  (let* ((db           #f)
	 (keys         (rmt:get-keys))
	 (rundat       (mt:get-runs-by-patt keys runnamepatt target))
	 (header       (vector-ref rundat 0))
	 (runs         (vector-ref rundat 1))
	 (states       (if state  (string-split state  ",") '()))

			     (if (not (null? tal))
				 (loop (car tal)(cdr tal))))
			   (let* ((item-path     (db:test-get-item-path new-test-dat))
				  (test-name     (db:test-get-testname new-test-dat))
				  (run-dir       ;;(filedb:get-path *fdb*
				   ;; (rmt:sdb-qry 'getid 
				   (db:test-get-rundir new-test-dat)) ;; )    ;; run dir is from the link tree
				  (real-dir      (if (file-exists? run-dir)
						     (resolve-pathname run-dir)
						     #f))
				  (test-state    (db:test-get-state new-test-dat))
				  (test-fulln    (db:test-get-fullname new-test-dat)))
				  (test-fulln    (db:test-get-fullname new-test-dat))
				  (uname         (db:test-get-uname    new-test-dat))
				  (toplevel-with-children (and (db:test-get-is-toplevel test)
							       (> (cdb:remote-run db:test-toplevel-num-items db run-id test-name) 0))))
			     (case action
			       ((remove-runs)
				;; if the test is a toplevel-with-children issue an error and do not remove
				(if toplevel-with-children
				    (debug:print 0 "WARNING: skipping removal of " test-fulln " with run-id " run-id " as it has sub tests")
				    (begin
				(debug:print-info 0 "test: " test-name " itest-state: " test-state)
				(if (member test-state (list "RUNNING" "LAUNCHED" "REMOTEHOSTSTART" "KILLREQ"))
				    (begin
				      (if (not (hash-table-ref/default test-retry-time test-fulln #f))
					  (begin
					    ;; want to set to REMOVING BUT CANNOT do it here?
					    (hash-table-set! test-retry-time test-fulln (current-seconds))))
				      (if (> (- (current-seconds)(hash-table-ref test-retry-time test-fulln)) allow-run-time)
					  ;; This test is not in a correct state for cleaning up. Let's try some graceful shutdown steps first
					  ;; Set the test to "KILLREQ" and wait five seconds then try again. Repeat up to five times then give
					  ;; up and blow it away.
					  (begin
					    (debug:print 0 "WARNING: could not gracefully remove test " test-fulln ", tried to kill it to no avail. Forcing state to FAILEDKILL and continuing")
				      (debug:print-info 0 "test: " test-name " itest-state: " test-state)
				      (if (member test-state (list "RUNNING" "LAUNCHED" "REMOTEHOSTSTART" "KILLREQ"))
					  (begin
					    (if (not (hash-table-ref/default test-retry-time test-fulln #f))
						(begin
						  ;; want to set to REMOVING BUT CANNOT do it here?
						  (hash-table-set! test-retry-time test-fulln (current-seconds))))
					    (if (> (- (current-seconds)(hash-table-ref test-retry-time test-fulln)) allow-run-time)
						;; This test is not in a correct state for cleaning up. Let's try some graceful shutdown steps first
						;; Set the test to "KILLREQ" and wait five seconds then try again. Repeat up to five times then give
						;; up and blow it away.
						(begin
						  (debug:print 0 "WARNING: could not gracefully remove test " test-fulln ", tried to kill it to no avail. Forcing state to FAILEDKILL and continuing")
					    (mt:test-set-state-status-by-id run-id (db:test-get-id test) "FAILEDKILL" "n/a" #f)
					    (thread-sleep! 1))
					  (begin
						  (thread-sleep! 1))
						(begin
					    (mt:test-set-state-status-by-id run-id (db:test-get-id test) "KILLREQ" "n/a" #f)
					    (thread-sleep! 1)))
				      ;; NOTE: This is suboptimal as the testdata will be used later and the state/status may have changed ...
				      (if (null? tal)
					  (loop new-test-dat tal)
					  (loop (car tal)(append tal (list new-test-dat)))))
				    (begin
						  (thread-sleep! 1)))
					    ;; NOTE: This is suboptimal as the testdata will be used later and the state/status may have changed ...
					    (if (null? tal)
						(loop new-test-dat tal)
						(loop (car tal)(append tal (list new-test-dat)))))
					  (begin
				      (mt:test-set-state-status-by-id run-id (db:test-get-id test) "REMOVING" "LOCKED" #f)
				      (debug:print-info 1 "Attempting to remove " (if real-dir (conc " dir " real-dir " and ") "") " link " run-dir)
				      (if (and real-dir 
					       (> (string-length real-dir) 5)
					       (file-exists? real-dir)) ;; bad heuristic but should prevent /tmp /home etc.
					  (begin ;; let* ((realpath (resolve-pathname run-dir)))
					    (debug:print-info 1 "Recursively removing " real-dir)
					    (if (file-exists? real-dir)
						(runs:safe-delete-test-dir real-dir)
						(debug:print 0 "WARNING: test dir " real-dir " appears to not exist or is not readable")))
					  (if real-dir 
					      (debug:print 0 "WARNING: directory " real-dir " does not exist")
					      (debug:print 0 "WARNING: no real directory corrosponding to link " run-dir ", nothing done")))
				      (if (symbolic-link? run-dir)
					  (begin
					    (debug:print-info 1 "Removing symlink " run-dir)
					    (handle-exceptions
					     exn
					     (debug:print 0 "ERROR:  Failed to remove symlink " run-dir ((condition-property-accessor 'exn 'message) exn) ", attempting to continue")
					     (delete-file run-dir)))
					  (if (directory? run-dir)
					      (if (> (directory-fold (lambda (f x)(+ 1 x)) 0 run-dir) 0)
						  (debug:print 0 "WARNING: refusing to remove " run-dir " as it is not empty")
						  (handle-exceptions
						   exn
						   (debug:print 0 "ERROR:  Failed to remove directory " run-dir ((condition-property-accessor 'exn 'message) exn) ", attempting to continue")
						   (delete-directory run-dir)))
					    (runs:remove-test-directory db new-test-dat remove-data-only)
					      (if run-dir
						  (debug:print 0 "WARNING: not removing " run-dir " as it either doesn't exist or is not a symlink")
						  (debug:print 0 "NOTE: the run dir for this test is undefined. Test may have already been deleted."))
					      ))
				      ;; Only delete the records *after* removing the directory. If things fail we have a record 
				      (rmt:delete-test-records (db:test-get-run_id test)(db:test-get-id test))
				      (if (not (null? tal))
					  (loop (car tal)(cdr tal))))))
					    (if (not (null? tal))
						(loop (car tal)(cdr tal))))))))
			       ((set-state-status)
				(debug:print-info 2 "new state " (car state-status) ", new status " (cadr state-status))
				(mt:test-set-state-status-by-id run-id (db:test-get-id test) (car state-status)(cadr state-status) #f)
				(if (not (null? tal))
				    (loop (car tal)(cdr tal))))
			       ((run-wait)
				(debug:print-info 2 "still waiting, " (length tests) " tests still running")

		       ;; 	 (debug:print 1 "Removing run dir " runpath)
		       ;; 	 (system (conc "rmdir -p " runpath))))
		       )))))
	 ))
     runs))
  #t)

(define (runs:remove-test-directory db test remove-data-only)
  (let* ((run-dir       (db:test-get-rundir test))    ;; run dir is from the link tree
	 (real-dir      (if (file-exists? run-dir)
			    (resolve-pathname run-dir)
			    #f)))
    (if (not remove-data-only)
	(mt:test-set-state-status-by-id (db:test-get-id test) "REMOVING" "LOCKED" #f))
    (debug:print-info 1 "Attempting to remove " (if real-dir (conc " dir " real-dir " and ") "") " link " run-dir)
    (if (and real-dir 
	     (> (string-length real-dir) 5)
	     (file-exists? real-dir)) ;; bad heuristic but should prevent /tmp /home etc.
	(begin ;; let* ((realpath (resolve-pathname run-dir)))
	  (debug:print-info 1 "Recursively removing " real-dir)
	  (if (file-exists? real-dir)
	      (runs:safe-delete-test-dir real-dir)
	      (debug:print 0 "WARNING: test dir " real-dir " appears to not exist or is not readable")))
	(if real-dir 
	    (debug:print 0 "WARNING: directory " real-dir " does not exist")
	    (debug:print 0 "WARNING: no real directory corrosponding to link " run-dir ", nothing done")))
    (if (symbolic-link? run-dir)
	(begin
	  (debug:print-info 1 "Removing symlink " run-dir)
	  (handle-exceptions
	   exn
	   (debug:print 0 "ERROR:  Failed to remove symlink " run-dir ((condition-property-accessor 'exn 'message) exn) ", attempting to continue")
	   (delete-file run-dir)))
	(if (directory? run-dir)
	    (if (> (directory-fold (lambda (f x)(+ 1 x)) 0 run-dir) 0)
		(debug:print 0 "WARNING: refusing to remove " run-dir " as it is not empty")
		(handle-exceptions
		 exn
		 (debug:print 0 "ERROR:  Failed to remove directory " run-dir ((condition-property-accessor 'exn 'message) exn) ", attempting to continue")
		 (delete-directory run-dir)))
	    (if run-dir
		(debug:print 0 "WARNING: not removing " run-dir " as it either doesn't exist or is not a symlink")
		(debug:print 0 "NOTE: the run dir for this test is undefined. Test may have already been deleted."))
	    ))
    ;; Only delete the records *after* removing the directory. If things fail we have a record 
    (if (not remove-data-only)
	(cdb:remote-run db:delete-test-records db #f (db:test-get-id test)))))

;;======================================================================
;; Routines for manipulating runs
;;======================================================================

;; Since many calls to a run require pretty much the same setup 
;; this wrapper is used to reduce the replication of code
(define (general-run-call switchname action-desc proc)