Megatest

(define (open-db) ;;  (conc *toppath* "/megatest.db") (car *configinfo*)))
  (if (not *toppath*)
      (if (not (setup-for-run))
	  (begin
	    (debug:print 0 "ERROR: Attempted to open db when not in megatest area. Exiting.")
	    (exit))))
  (let* ((dbpath       (conc *toppath* "/megatest.db")) ;; fname)
	 (dbexists     (file-exists? dbpath))
	 (write-access (file-write-access? dbpath))
	 (db           (sqlite3:open-database dbpath)) ;; (never-give-up-open-db dbpath))
	 (handler      (make-busy-timeout (if (args:get-arg "-override-timeout")
					      (string->number (args:get-arg "-override-timeout"))
					      6000)))) ;; NB// this is in milliseconds. 136000))) ;; 136000 = 2.2 minutes
    (if (and dbexists
	     (not write-access))
	(set! *db-write-access* write-access)) ;; only unset so other db's also can use this control
    (debug:print-info 11 "open-db, dbpath=" dbpath " argv=" (argv))
    (if write-access (sqlite3:set-busy-handler! db handler))
    (if (not dbexists)
	(db:initialize db))

		(debug:print-info 4 "Trying " (car pathdat) " at " (cadr pathdat))
		(sqlite3:finalize! db)
		(if (not (null? newres))
		    (car newres)
		    (if (null? tal)
			#f
			(loop (car tal)(cdr tal))))))))))

(define (db:test-toplevel-num-items db run-id testname)
  (let ((res 0))
    (sqlite3:for-each-row
     (lambda (num-items)
       (set! res num-items))
     db
     "SELECT count(id) FROM tests WHERE run_id=? AND testname=? AND item_path != '' AND state NOT IN ('DELETED');"
     run-id
     testname)
    res))

;;======================================================================
;; QUEUE UP META, TEST STATUS AND STEPS REMOTE ACCESS
;;======================================================================

;; NOTE: Can remove the regex and base64 encoding for zmq
(define (db:obj->string obj)

;; Test record accessors
;;
(define (make-db:test)(make-vector 20))
(define-inline (db:test-get-id           vec) (vector-ref vec 0))
(define-inline (db:test-get-run_id       vec) (vector-ref vec 1))
(define-inline (db:test-get-testname     vec) (vector-ref vec 2))
(define-inline (db:test-get-state        vec) (vector-ref vec 3))
(define-inline (db:test-get-status       vec) (vector-ref vec 4))
(define-inline (db:test-get-event_time   vec) (vector-ref vec 5))

(define-inline (db:test-set-diskfree! vec val)(vector-set! vec 8 val))
(define-inline (db:test-set-testname! vec val)(vector-set! vec 2 val))
(define-inline (db:test-set-state!    vec val)(vector-set! vec 3 val))
(define-inline (db:test-set-status!   vec val)(vector-set! vec 4 val))
(define-inline (db:test-set-run_duration! vec val)(vector-set! vec 12 val))
(define-inline (db:test-set-final_logf! vec val)(vector-set! vec 13 val))

;; Test record utility functions

;; Is a test a toplevel?
;;
(define (db:test-get-is-toplevel vec)
  (and (equal? (db:test-get-item-path vec) "")      ;; test is not an item
       (equal? (db:test-get-uname vec)     "n/a"))) ;; test has never been run

;; get rows and header from 
(define-inline (db:get-header vec)(vector-ref vec 0))
(define-inline (db:get-rows   vec)(vector-ref vec 1))

;; make-vector-record "" db mintest id run_id testname state status event_time item_path
;;
(define (make-db:mintest)(make-vector 7))

      (if (not (setup-for-run))
	  (begin
	    (debug:print 0 "Failed to setup, exiting") 
	    (exit 1)))
      (open-run-close db:find-and-mark-incomplete #f)
      (set! *didsomething* #t)))














;;======================================================================
;; Update the tests meta data from the testconfig files
;;======================================================================

(if (args:get-arg "-update-meta")
    (begin
      (if (not (setup-for-run))

	      (or (get-environment-variable "HOME") ".") "/.megatest_history"))
	    (current-input-port (make-gnu-readline-port "megatest> "))
	    (if (args:get-arg "-repl")
		(repl)
		(load (args:get-arg "-load"))))
	  (exit))
      (set! *didsomething* #t)))

;;======================================================================
;; Wait on a run to complete
;;======================================================================

(if (args:get-arg "-run-wait")
    (begin
      (if (not (setup-for-run))
	  (begin
	    (debug:print 0 "Failed to setup, exiting") 
	    (exit 1)))
      (operate-on 'run-wait)
      (set! *didsomething* #t)))

;;======================================================================
;; Exit and clean up
;;======================================================================

(if *runremote* (close-all-connections!))

			   (let* ((item-path     (db:test-get-item-path new-test-dat))
				  (test-name     (db:test-get-testname new-test-dat))
				  (run-dir       (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))
				  (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")
						  (mt:test-set-state-status-by-id (db:test-get-id test) "FAILEDKILL" "n/a" #f)
						  (thread-sleep! 1))
						(begin
						  (mt:test-set-state-status-by-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
					    (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)))
					    (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 (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")