Megatest

tests.o runs.o dashboard.o dashboard-tests.o dashboard-main.o  : run_records.scm
db.o ezsteps.o keys.o launch.o megatest.o monitor.o runs-for-ref.o runs.o tests.o : key_records.scm
tests.o tasks.o dashboard-tasks.o : task_records.scm
runs.o : test_records.scm
megatest.o : megatest-fossil-hash.scm

# Temporary while transitioning to new routine
runs.o : run-tests-queue-classic.scm  run-tests-queue-new.scm

megatest-fossil-hash.scm : $(SRCFILES) megatest.scm *_records.scm
	echo "(define megatest-fossil-hash \"$(MTESTHASH)\")" > megatest-fossil-hash.new
	if ! diff -q megatest-fossil-hash.new megatest-fossil-hash.scm ; then echo copying .new to .scm;cp -f megatest-fossil-hash.new megatest-fossil-hash.scm;fi

$(OFILES) $(GOFILES) : common_records.scm 

1.  Run all tests
2.  create run areas, copy in conf and scripts                   DONE
3.  Add a host chooser for ssh to launch-tests
4.  Run creation timestamp not happening                         DONE
5 . Check for test already in progress, give meaningful message  DONE
6.  Debug xterm creation for test generation                     DONE
7.  Capture run info, host, load, freemem at test launch         DONE
8.  Rename to testalot?                                          Nah! I like Megatest
10. Run, test and step comment field                             
11. At end of test scan all tests for this run, if all done 
    update run status to COMPLETED                               NOT gonna happen. It is up to the test to mark as PASS/FAIL
12. state and status lists need to be regexes                    
13. Test on Chicken 4.                                           DONE
14. Try making static executable
15. Log processor script                                         DONE

(define *keys*              (make-hash-table)) ;; cache the keys here
(define *keyvals*           (make-hash-table))
(define *toptest-paths*     (make-hash-table)) ;; cache toptest path settings here
(define *test-paths*        (make-hash-table)) ;; cache test-id to test run paths here
(define *test-ids*          (make-hash-table)) ;; cache run-id, testname, and item-path => test-id
(define *test-info*         (make-hash-table)) ;; cache the test info records, update the state, status, run_duration etc. from testdat.db



;; Awful. Please FIXME
(define *env-vars-by-run-id* (make-hash-table))
(define *current-run-name*   #f)

(define (common:clear-caches)

  (set! *keys*               (make-hash-table))
  (set! *keyvals*            (make-hash-table))
  (set! *toptest-paths*      (make-hash-table))
  (set! *test-paths*         (make-hash-table))
  (set! *test-ids*           (make-hash-table))
  (set! *test-info*          (make-hash-table))

  (set! *env-vars-by-run-id* (make-hash-table))
  (set! *test-id-cache*      (make-hash-table)))

;; Debugging stuff
(define *verbosity*         1)
(define *logging*           #f)

			       #f #t))) string<?))

;; '(print (string-intersperse (map cadr (hash-table-ref/default (read-config "megatest.config" \#f \#t) "disks" '"'"'("none" ""))) "\n"))'
(define (common:get-disks)
  (hash-table-ref/default 
   (read-config "megatest.config" #f #t)
   "disks" '("none" "")))



























;;======================================================================
;; System stuff
;;======================================================================

;; return a nice clean pathname made absolute
(define (nice-path dir)

											      (db:test-data-get-units    x)
											      (db:test-data-get-type     x)
											      (db:test-data-get-comment  x)))
										    (open-run-close db:read-test-data #f test-id "%")))
									      "\n")))
							       (if (not (equal? currval newval))
								   (iup:attribute-set! test-data "VALUE" newval ))))) ;; "TITLE" newval)))))
					  test-data)))))


				 (iup:attribute-set! tabs "TABTITLE0" "Steps")
				 (iup:attribute-set! tabs "TABTITLE1" "Test Data")
				 tabs))))
	    (iup:show self)
	    (iup:callback-set! *tim* "ACTION_CB"
			       (lambda (x)
				 ;; Now start keeping the gui updated from the db

;;======================================================================

(use format)
(require-library iup)
(import (prefix iup iup:))

(use canvas-draw)


(use sqlite3 srfi-1 posix regex regex-case srfi-69)
(import (prefix sqlite3 sqlite3:))


(declare (uses common))
(declare (uses margs))
(declare (uses keys))
(declare (uses items))
(declare (uses db))
(declare (uses configf))

(define (update-search x val)
  ;; (print "Setting search for " x " to " val)
  (hash-table-set! *searchpatts* x val))

(define (mark-for-update)
  (set! *last-db-update-time* 0)
  (set! *delayed-update* 1)
  )














































































































































































































(define (make-dashboard-buttons nruns ntests keynames)
  (let* ((nkeys   (length keynames))
	 (runsvec (make-vector nruns))
	 (header  (make-vector nruns))
	 (lftcol  (make-vector ntests))
	 (keycol  (make-vector ntests))

	  (hash-table-set! *buttondat* button-key (vector 0 "100 100 100" button-key #f #f)) 
	  (vector-set! testvec testnum butn)
	  (loop runnum (+ testnum 1) testvec (cons butn res))))))
    ;; now assemble the hdrlst and bdylst and kick off the dialog
    (iup:show
     (iup:dialog 
      #:title "Megatest dashboard"

      (iup:vbox
	(apply iup:hbox 
	       (cons (apply iup:vbox lftlst)
		     (list 
		      (iup:vbox
		       ;; the header
		       (apply iup:hbox (reverse hdrlst))
		       (apply iup:hbox (reverse bdylst))))))
       controls)))





    (vector keycol lftcol header runsvec)))

(if (or (args:get-arg "-rows")
	(get-environment-variable "DASHBOARDROWS" ))
    (begin
        (set! *num-tests* (string->number (or (args:get-arg "-rows")
					      (get-environment-variable "DASHBOARDROWS"))))
	(update-rundat "%" *num-runs* "%/%" '()))

	 (lambda (key)
	   (set! res (cons key res)))
	 db
	 "SELECT fieldname FROM keys ORDER BY id DESC;")
	(set! *db-keys* res)
	res)))


(define (db:get-value-by-header row header field)
  (debug:print-info 4 "db:get-value-by-header row: " row " header: " header " field: " field)
  (if (null? header) #f
      (let loop ((hed (car header))
		 (tal (cdr header))
		 (n   0))
	(if (equal? hed field)

       (set! res (cons (apply vector a x) res)))
     db
     qrystr
     )
    (debug:print-info 11 "db:get-runs END qrystr: " qrystr " keypatts: " keypatts " offset: " offset " limit: " count)
    (vector header res)))






















;; just get count of runs
(define (db:get-num-runs db runpatt)
  (let ((numruns 0))
    (debug:print-info 11 "db:get-num-runs START " runpatt)
    (sqlite3:for-each-row 
     (lambda (count)
       (set! numruns count))
     db
     "SELECT COUNT(id) FROM runs WHERE runname LIKE ? AND state != 'DELETED';" runpatt)
    (debug:print-info 11 "db:get-num-runs END " runpatt)
    numruns))






































;; use (get-value-by-header (db:get-header runinfo)(db:get-row runinfo))
(define (db:get-run-info db run-id)
  ;;(if (hash-table-ref/default *run-info-cache* run-id #f)
  ;;    (hash-table-ref *run-info-cache* run-id)
      (let* ((res      #f)
	     (keys      (db:get-keys db))
	     (remfields (list "id" "runname" "state" "status" "owner" "event_time"))

    (reverse res)))

;; get key vals for a given run-id
(define (db:get-key-vals db run-id)
   (let ((mykeyvals (hash-table-ref/default *keyvals* run-id #f)))
    (if mykeyvals 
	mykeyvals
  (let* ((keys (db:get-keys db))
	 (res  '()))
    (debug:print-info 11 "db:get-key-vals START keys: " keys " run-id: " run-id)
    (for-each 
     (lambda (key)
       (let ((qry (conc "SELECT " key " FROM runs WHERE id=?;")))
	 ;; (debug:print 0 "qry: " qry)
	 (sqlite3:for-each-row 

The Megatest Users Manual
=========================
Matt Welland <matt@kiatoa.com>
v1.0, April 2012
:doctype: book


[dedication]
Dedication
==========

Dedicated to my wife Joanna who has kindly supported my working on various projects over the years.

Thanks
------

Thank you the many people I've worked over the years who have 
shared their knowledge and insights with me.

Thanks also to the creators of the various open source projects that
Megatest is built on. These include Linux, xemacs, chicken scheme,
fossil and asciidoc. Without these projects something like Megatest 
would be difficult or impossible to do.

[preface]
Preface
=======
This book is organised as three sub-books; getting started, writing tests and reference.

Why Megatest?
~~~~~~~~~~~~~

			    (append (list (list "MT_TEST_RUN_DIR" work-area)
					  (list "MT_TEST_NAME" test-name)
					  (list "MT_ITEM_INFO" (conc itemdat)) 
					  (list "MT_RUNNAME"   runname)
					  (list "MT_TARGET"    mt_target)
					  )
				    itemdat)))

	   (launch-results (apply (if (equal? (configf:lookup *configdat* "setup" "launchwait") "yes")

				      cmd-run-with-stderr->list
				      process-run)
				  (if useshell
				      (string-intersperse fullcmd " ")
				      (car fullcmd))
				  (if useshell
				      '()
				      (cdr fullcmd)))))
      (if (list? launch-results)

	  (with-output-to-file "mt_launch.log"
	    (lambda ()

	      (apply print launch-results))

	    #:append))
      (debug:print 2 "Launching completed, updating db")
      (debug:print 2 "Launch results: " launch-results)
      (if (not launch-results)
          (begin
            (print "ERROR: Failed to run " (string-intersperse fullcmd " ") ", exiting now")
            ;; (sqlite3:finalize! db)
            ;; good ole "exit" seems not to work

(define (args:get-arg arg . default)
  (if (null? default)
      (hash-table-ref/default args:arg-hash arg #f)
      (hash-table-ref/default args:arg-hash arg (car default))))

(define (args:get-arg-from ht arg . default)
  (if (null? default)
      (hash-table-ref/default ht arg #f))
      (hash-table-ref/default ht arg (car default)))

(define (args:usage . args)
  (if (> (length args) 0)
      (apply print "ERROR: " args))
  (if (string? help)
      (print help)
      (print "Usage: " (car (argv)) " ... "))
  (exit 0))

;; args: 
(define (args:get-args args params switches arg-hash num-needed)
  (let* ((numargs (length args))
	 (adj-num-needed (if num-needed (+ num-needed 2) #f)))
    (if (< numargs (if adj-num-needed adj-num-needed 2))
	(if (>= num-needed 1)
	    (args:usage "No arguments provided"))

	(let loop ((arg (cadr args))
		   (tail (cddr args))
		   (remargs '()))
	  (cond 
	   ((member arg params) ;; args with params
	    (if (< (length tail) 1)
		(args:usage "param given without argument " arg)

;; Always use two digit decimal
;; 1.01, 1.02...1.10,1.11 ... 1.99,2.00..

(declare (unit megatest-version))

(define megatest-version 1.5504)

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

;; This is the Megatest API. All generally "useful" routines will be wrapped or extended
;; here.



















(define (mt:get-tests-for-run run-id testpatt states status #!key (not-in #t) (sort-by #f))
  (let loop ((testsdat (cdb:remote-run db:get-tests-for-run #f run-id testpatt states status 0 500 not-in sort-by))
	     (res      '())
	     (offset   0)
	     (limit    500))
    (let* ((full-list (append res testsdat))

;; test-records is a hash table testname:item_path => vector < testname testconfig waitons priority items-info ... >
(define (runs:run-tests-queue-classic run-id runname test-records keyvals flags test-patts required-tests)
    ;; At this point the list of parent tests is expanded 
    ;; NB// Should expand items here and then insert into the run queue.
  (debug:print 5 "test-records: " test-records ", flags: " (hash-table->alist flags))
  (let ((run-info              (cdb:remote-run db:get-run-info #f run-id))
	(sorted-test-names     (tests:sort-by-priority-and-waiton test-records))
	(test-registry         (make-hash-table))
	(registry-mutex        (make-mutex))
	(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))))
    (set! max-retries (if (and max-retries (string->number max-retries))(string->number max-retries) 100))
    (if (not (null? sorted-test-names))
	(let loop ((hed         (car sorted-test-names))
		   (tal         (cdr sorted-test-names))
		   (reruns      '()))
	  (if (not (null? reruns))(debug:print-info 4 "reruns=" reruns))
	  ;; (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 "requirements" "jobgroup"))
		 (testmode    (let ((m (config-lookup tconfig "requirements" "mode")))
				(if m (string->symbol m) 'normal)))
		 (waitons     (tests:testqueue-get-waitons    test-record))
		 (priority    (tests:testqueue-get-priority   test-record))
		 (itemdat     (tests:testqueue-get-itemdat    test-record)) ;; itemdat can be a string, list or #f
		 (items       (tests:testqueue-get-items      test-record))
		 (item-path   (item-list->path itemdat))
		 (newtal      (append tal (list hed))))
	    
	    (debug:print 6
			 "test-name: " test-name
			 "\n  hed:         " hed
			 "\n  itemdat:     " itemdat
			 "\n  items:       " items
			 "\n  item-path:   " item-path
			 "\n  waitons:     " waitons
			 "\n  num-retries: " num-retries
			 "\n  tal:         " tal
			 "\n  reruns:      " reruns)

	    ;; check for hed in waitons => this would be circular, remove it and issue an
	    ;; error
	    (if (member test-name waitons)
		(begin
		  (debug:print 0 "ERROR: test " test-name " has listed itself as a waiton, please correct this!")
		  (set! waiton (filter (lambda (x)(not (equal? x hed))) waitons))))

	    (cond ;; OUTER COND
	     ((not items) ;; when false the test is ok to be handed off to launch (but not before)
	      (if (and (not (tests:match test-patts (tests:testqueue-get-testname test-record) item-path required: required-tests))
	               (not (null? tal)))
	          (loop (car newtal)(cdr newtal) reruns))
	      (let* ((run-limits-info         ;; (cdb:remote-run runs:can-run-more-tests #f jobgroup max-concurrent-jobs)) ;; look at the test jobgroup and tot jobs running
		      (open-run-close runs:can-run-more-tests #f jobgroup max-concurrent-jobs)) ;; look at the test jobgroup and tot jobs running
		     (have-resources          (car run-limits-info))
		     (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         (db:get-prereqs-not-met run-id waitons item-path mode: testmode))
		     (fails                   (runs:calc-fails prereqs-not-met))
		     (non-completed           (runs:calc-not-completed prereqs-not-met)))
		(debug:print-info 8 "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)
		(debug:print-info 4 "hed=" hed "\n  test-record=" test-record "\n  test-name: " test-name "\n  item-path: " item-path "\n  test-patts: " test-patts)

		;; 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?
		(debug:print-info 4 "run-limits-info = " run-limits-info)
		(cond ;; INNER COND #1 for a launchable test
		 ;; Check item path against item-patts
		 ((not (tests:match test-patts (tests:testqueue-get-testname test-record) item-path required: required-tests)) ;; This test/itempath is not to be run
		  ;; else the run is stuck, temporarily or permanently
		  ;; but should check if it is due to lack of resources vs. prerequisites
		  (debug:print-info 1 "Skipping " (tests:testqueue-get-testname test-record) " " item-path " as it doesn't match " test-patts)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (car tal)(cdr tal) reruns)))
		 ;; Registry has been started for this test but has not yet completed
		 ;; this should be rare, the case where there are only a couple of tests and the db is slow
		 ;; delay a short while and continue
		 ;; ((eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f) 'start)
		 ;;  (thread-sleep! 0.01)
		 ;;  (loop (car newtal)(cdr newtal) reruns))
		 ;; count number of 'done, if more than 100 then skip on through.
		 ((not (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f)) ;; ) ;; too many changes required. Implement later.
		  (debug:print-info 4 "Pre-registering test " test-name "/" item-path " to create placeholder" )
		  (let ((th (make-thread (lambda ()
		        		   (mutex-lock! registry-mutex)
		        		   (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'start)
		        		   (mutex-unlock! registry-mutex)
					   ;; If haven't done it before register a top level test if this is an itemized test
					   (if (not (eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name "") #f) 'done))
					       (cdb:tests-register-test *runremote* run-id test-name ""))
					   (cdb:tests-register-test *runremote* run-id test-name item-path)
		        		   (mutex-lock! registry-mutex)
					   (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'done)
		        		   (mutex-unlock! registry-mutex))
		        		 (conc test-name "/" item-path))))
		    (thread-start! th))
		  (runs:shrink-can-run-more-tests-count)   ;; DELAY TWEAKER (still needed?)
		  (loop (car newtal)(cdr newtal) reruns))
		 ;; At this point *all* test registrations must be completed.
		 ((not (null? (filter (lambda (x)(eq? 'start x))(hash-table-values test-registry))))
		  (debug:print-info 0 "Waiting on test registrations: " (string-intersperse 
									 (filter (lambda (x)
										   (eq? (hash-table-ref/default test-registry x #f) 'start))
										 (hash-table-keys test-registry))
									 ", "))
		  (thread-sleep! 0.1)
		  (loop hed tal reruns))
		 ((not have-resources) ;; simply try again after waiting a second
		  (debug:print-info 1 "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! 1) ;; (+ 2 *global-delta*))
		  ;; could have done hed tal here but doing car/cdr of newtal to rotate tests
		  (loop (car newtal)(cdr newtal) reruns))
		 ((and have-resources
		       (or (null? prereqs-not-met)
			   (and (eq? testmode 'toplevel)
				(null? non-completed))))
		  (run:test run-id run-info keyvals runname test-record flags #f)
		  (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'running)
		  (runs:shrink-can-run-more-tests-count)  ;; DELAY TWEAKER (still needed?)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (car tal)(cdr tal) reruns)))
		 (else ;; must be we have unmet prerequisites
		    (debug:print 4 "FAILS: " fails)
		    ;; If one or more of the prereqs-not-met are FAIL then we can issue
		    ;; a message and drop hed from the items to be processed.
		    (if (null? fails)
			(begin
			  ;; couldn't run, take a breather
			  (debug:print-info 4 "Shouldn't really get here, race condition? Unable to launch more tests at this moment, killing time ...")
			  ;; (thread-sleep! (+ 0.01 *global-delta*)) ;; long sleep here - no resources, may as well be patient
			  ;; we made new tal by sticking hed at the back of the list
			  (loop (car newtal)(cdr newtal) reruns))
			;; the waiton is FAIL so no point in trying to run hed ever again
			(if (not (null? tal))
			    (if (vector? hed)
				(begin 
				  (debug:print 1 "WARN: Dropping test " (db:test-get-testname hed) "/" (db:test-get-item-path hed)
					       " from the launch list as it has prerequistes that are FAIL")
				  (runs:shrink-can-run-more-tests-count) ;; DELAY TWEAKER (still needed?)
				  ;; (thread-sleep! *global-delta*)
				  (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'removed)
				  (loop (car tal)(cdr tal) (cons hed reruns)))
				(begin
				  (debug:print 1 "WARN: Test not processed correctly. Could be a race condition in your test implementation? " hed) ;;  " as it has prerequistes that are FAIL. (NOTE: hed is not a vector)")
				  (runs:shrink-can-run-more-tests-count) ;; DELAY TWEAKER (still needed?)
				  ;; (thread-sleep! (+ 0.01 *global-delta*))
				  (loop hed tal reruns))))))))) ;; END OF INNER COND
	     
	     ;; case where an items came in as a list been processed
	     ((and (list? items)     ;; thus we know our items are already calculated
		   (not   itemdat)) ;; and not yet expanded into the list of things to be done
	      (if (and (debug:debug-mode 1) ;; (>= *verbosity* 1)
		       (> (length items) 0)
		       (> (length (car items)) 0))
		  (pp items))
	      (for-each
	       (lambda (my-itemdat)
		 (let* ((new-test-record (let ((newrec (make-tests:testqueue)))
					   (vector-copy! test-record newrec)
					   newrec))
			(my-item-path (item-list->path my-itemdat)))
		   (if (tests:match test-patts hed my-item-path required: required-tests) ;; (patt-list-match my-item-path item-patts)           ;; yes, we want to process this item, NOTE: Should not need this check here!
		       (let ((newtestname (runs:make-full-test-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)
			 (set! tal (cons newtestname tal)))))) ;; since these are itemized create new test names testname/itempath
	       items)
	      (if (not (null? tal))
		  (begin
		    (debug:print-info 4 "End of items list, looping with next after short delay")
                    ;; (thread-sleep! (+ 0.01 *global-delta*))
		    (loop (car tal)(cdr tal) reruns))))

	     ;; if items is a proc then need to run items:get-items-from-config, get the list and loop 
	     ;;    - but only do that if resources exist to kick off the job
	     ((or (procedure? items)(eq? items 'have-procedure))
	      (let ((can-run-more    ;; (cdb:remote-run runs:can-run-more-tests #f jobgroup max-concurrent-jobs)))
		     (open-run-close runs:can-run-more-tests #f jobgroup max-concurrent-jobs)))
		(if (and (list? can-run-more)
			 (car can-run-more))
		    (let* ((prereqs-not-met (db:get-prereqs-not-met run-id waitons item-path mode: testmode))
			   (fails           (runs:calc-fails prereqs-not-met))
			   (non-completed   (runs:calc-not-completed prereqs-not-met)))
		      (debug:print-info 8 "can-run-more: " can-run-more
				   "\n testname:        " hed
				   "\n prereqs-not-met: " (runs:pretty-string prereqs-not-met)
				   "\n non-completed:   " (runs:pretty-string non-completed) 
				   "\n fails:           " (runs:pretty-string fails)
				   "\n testmode:        " testmode
				   "\n num-retries:     " num-retries
				   "\n (eq? testmode 'toplevel): " (eq? testmode 'toplevel)
				   "\n (null? non-completed):    " (null? non-completed)
				   "\n reruns:          " reruns
				   "\n items:           " items
				   "\n can-run-more:    " can-run-more)
		      ;; (thread-sleep! (+ 0.01 *global-delta*))
		      (cond ;; INNER COND #2
		       ((or (null? prereqs-not-met) ;; all prereqs met, fire off the test
			    ;; or, if it is a 'toplevel test and all prereqs not met are COMPLETED then launch
			    (and (eq? testmode 'toplevel)
				 (null? non-completed)))
			(let ((test-name (tests:testqueue-get-testname test-record)))
			  (setenv "MT_TEST_NAME" test-name) ;; 
			  (setenv "MT_RUNNAME"   runname)
			  (set-megatest-env-vars run-id inrunname: runname) ;; these may be needed by the launching process
			  (let ((items-list (items:get-items-from-config tconfig)))
			    (if (list? items-list)
				(begin
				  (tests:testqueue-set-items! test-record items-list)
				  ;; (thread-sleep! *global-delta*)
				  (loop hed tal reruns))
				(begin
				  (debug:print 0 "ERROR: The proc from reading the setup did not yield a list - please report this")
				  (exit 1))))))
		       ((null? fails)
			(debug:print-info 4 "fails is null, moving on in the queue but keeping " hed " for now")
			;; only increment num-retries when there are no tests runing
			(if (eq? 0 (list-ref can-run-more 1))
			    (begin
			      ;; TRY (if (> num-retries 100) ;; first 100 retries are low time cost
			      ;; TRY     (thread-sleep! (+ 2 *global-delta*))
			      ;; TRY     (thread-sleep! (+ 0.01 *global-delta*)))
			      (set! num-retries (+ num-retries 1))))
			(if (> num-retries  max-retries)
			    (if (not (null? tal))
				(loop (car tal)(cdr tal) reruns))
			    (loop (car newtal)(cdr newtal) reruns))) ;; an issue with prereqs not yet met?
		       ((and (not (null? fails))(eq? testmode 'normal))
			(debug:print-info 1 "test "  hed " (mode=" testmode ") has failed prerequisite(s); "
				     (string-intersperse (map (lambda (t)(conc (db:test-get-testname t) ":" (db:test-get-state t)"/"(db:test-get-status t))) fails) ", ")
				     ", removing it from to-do list")
			(if (not (null? tal))
			    (begin
                              ;; (thread-sleep! *global-delta*)
			      (loop (car tal)(cdr tal)(cons hed reruns)))))
		       (else
			(debug:print 8 "ERROR: No handler for this condition.")
			;; TRY (thread-sleep! (+ 1 *global-delta*))
			(loop (car newtal)(cdr newtal) reruns)))) ;; END OF IF CAN RUN MORE

		    ;; if can't run more just loop with next possible test
		    (begin
		      (debug:print-info 4 "processing the case with a lambda for items or 'have-procedure. Moving through the queue without dropping " hed)
		      ;; (thread-sleep! (+ 2 *global-delta*))
		      (loop (car newtal)(cdr newtal) reruns))))) ;; END OF (or (procedure? items)(eq? items 'have-procedure))
	     
	     ;; this case should not happen, added to help catch any bugs
	     ((and (list? items) itemdat)
	      (debug:print 0 "ERROR: Should not have a list of items in a test and the itemspath set - please report this")
	      (exit 1))
	     ((not (null? reruns))
	      (let* ((newlst (tests:filter-non-runnable run-id tal test-records)) ;; i.e. not FAIL, WAIVED, INCOMPLETE, PASS, KILLED,
		     (junked (lset-difference equal? tal newlst)))
		(debug:print-info 4 "full drop through, if reruns is less than 100 we will force retry them, reruns=" reruns ", tal=" tal)
		(if (< num-retries max-retries)
		    (set! newlst (append reruns newlst)))
		(set! num-retries (+ num-retries 1))
		;; (thread-sleep! (+ 1 *global-delta*))
		(if (not (null? newlst))
		    ;; since reruns have been tacked on to newlst create new reruns from junked
		    (loop (car newlst)(cdr newlst)(delete-duplicates junked)))))
	     ((not (null? tal))
	      (debug:print-info 4 "I'm pretty sure I shouldn't get here."))
	     (else
	      (debug:print-info 4 "Exiting loop with...\n  hed=" hed "\n  tal=" tal "\n  reruns=" reruns))
	     )))) ;; LET* ((test-record

    ;; we get here on "drop through" - loop for next test in queue
    ;; FIXME!!!! THIS SHOULD NOT REQUIRE AN EXIT!!!!!!!
    
    (debug:print-info 1 "All tests launched")
    (thread-sleep! 0.5)
    ;; FIXME! This harsh exit should not be necessary....
    ;; (if (not *runremote*)(exit)) ;; 
    #f)) ;; return a #f as a hint that we are done
  ;; Here we need to check that all the tests remaining to be run are eligible to run
  ;; and are not blocked by failed
  

;; test-records is a hash table testname:item_path => vector < testname testconfig waitons priority items-info ... >
(define (runs:run-tests-queue-new run-id runname test-records keyvals flags test-patts required-tests reglen-in)
  ;; At this point the list of parent tests is expanded 
  ;; NB// Should expand items here and then insert into the run queue.
  (debug:print 5 "test-records: " test-records ", flags: " (hash-table->alist flags))
  (let ((run-info              (cdb:remote-run db:get-run-info #f run-id))
	(tests-info            (mt:get-tests-for-run run-id #f '() '())) ;;  qryvals: "id,testname,item_path"))
	(sorted-test-names     (tests:sort-by-priority-and-waiton test-records))
	(test-registry         (make-hash-table))
	(registry-mutex        (make-mutex))
	(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)))
    ;; Initialize the test-registery hash with tests that already have a record
    (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)))
		  (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))
    (if (not (null? sorted-test-names))
	(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))
	  ;; (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 "requirements" "jobgroup"))
		 (testmode    (let ((m (config-lookup tconfig "requirements" "mode")))
				(if m (string->symbol m) 'normal)))
		 (waitons     (tests:testqueue-get-waitons    test-record))
		 (priority    (tests:testqueue-get-priority   test-record))
		 (itemdat     (tests:testqueue-get-itemdat    test-record)) ;; itemdat can be a string, list or #f
		 (items       (tests:testqueue-get-items      test-record))
		 (item-path   (item-list->path itemdat))
		 (tfullname   (runs:make-full-test-name test-name item-path))
		 (newtal      (append tal (list hed)))
		 (regfull     (> (length reg) reglen)))

	    ;; Fast skip of tests that are already "COMPLETED"
	    (if (equal? (hash-table-ref/default test-registry tfullname #f) 'COMPLETED)
		(begin
		  (debug:print-info 0 "Skipping COMPLETED test " tfullname)
		  (if (not (null? tal))
		      (loop (car tal)(cdr tal) reg reruns))))
	    ;; (if (> (length reg) 10)
	    ;;     (begin
	    ;;       (set! tal (cons hed tal))
	    ;;       (set! hed (car reg))
	    ;;       (set! reg (cdr reg))
	    ;;       (set! newtal tal)))
	    (debug:print 6
			 "test-name: " test-name
			 "\n  hed:         " hed
			 "\n  itemdat:     " itemdat
			 "\n  items:       " items
			 "\n  item-path:   " item-path
			 "\n  waitons:     " waitons
			 "\n  num-retries: " num-retries
			 "\n  tal:         " tal
			 "\n  reruns:      " reruns)

	    ;; check for hed in waitons => this would be circular, remove it and issue an
	    ;; error
	    (if (member test-name waitons)
		(begin
		  (debug:print 0 "ERROR: test " test-name " has listed itself as a waiton, please correct this!")
		  (set! waiton (filter (lambda (x)(not (equal? x hed))) waitons))))

	    (cond ;; OUTER COND
	     ((not items) ;; when false the test is ok to be handed off to launch (but not before)
	      (if (and (not (tests:match test-patts (tests:testqueue-get-testname test-record) item-path required: required-tests))
	               (not (null? tal)))
	          (loop (car tal)(cdr tal) reg reruns))
	      (let* ((run-limits-info         (cdb:remote-run runs:can-run-more-tests #f jobgroup max-concurrent-jobs)) ;; look at the test jobgroup and tot jobs running
		      ;; (open-run-close runs:can-run-more-tests #f jobgroup max-concurrent-jobs)) ;; look at the test jobgroup and tot jobs running
		     (have-resources          (car run-limits-info))
		     (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         (mt:get-prereqs-not-met run-id waitons item-path mode: testmode))
		     (fails                   (runs:calc-fails prereqs-not-met))
		     (non-completed           (runs:calc-not-completed prereqs-not-met)))
		(debug:print-info 8 "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)
		(debug:print-info 4 "hed=" hed "\n  test-record=" test-record "\n  test-name: " test-name "\n  item-path: " item-path "\n  test-patts: " test-patts)

		;; 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?
		(debug:print-info 4 "run-limits-info = " run-limits-info)
		(cond ;; INNER COND #1 for a launchable test
		 ;; Check item path against item-patts
		 ((not (tests:match test-patts (tests:testqueue-get-testname test-record) item-path required: required-tests)) ;; This test/itempath is not to be run
		  ;; else the run is stuck, temporarily or permanently
		  ;; but should check if it is due to lack of resources vs. prerequisites
		  (debug:print-info 1 "Skipping " (tests:testqueue-get-testname test-record) " " item-path " as it doesn't match " test-patts)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (runs:queue-next-hed tal reg reglen regfull)
			    (runs:queue-next-tal tal reg reglen regfull)
			    (runs:queue-next-reg tal reg reglen regfull)
			    reruns)))
		 ;; Registry has been started for this test but has not yet completed
		 ;; this should be rare, the case where there are only a couple of tests and the db is slow
		 ;; delay a short while and continue
		 ;; ((eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f) 'start)
		 ;;  (thread-sleep! 0.01)
		 ;;  (loop (car newtal)(cdr newtal) reruns))
		 ;; count number of 'done, if more than 100 then skip on through.
		 ((not (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f)) ;; ) ;; too many changes required. Implement later.
		  (debug:print-info 4 "Pre-registering test " test-name "/" item-path " to create placeholder" )
		  (let ((th (make-thread (lambda ()
		        		   (mutex-lock! registry-mutex)
		        		   (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'start)
		        		   (mutex-unlock! registry-mutex)
					   ;; If haven't done it before register a top level test if this is an itemized test
					   (if (not (eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name "") #f) 'done))
					       (cdb:tests-register-test *runremote* run-id test-name ""))
					   (cdb:tests-register-test *runremote* run-id test-name item-path)
		        		   (mutex-lock! registry-mutex)
					   (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'done)
		        		   (mutex-unlock! registry-mutex))
		        		 (conc test-name "/" item-path))))
		    (thread-start! th))
		  (cdb:remote-run runs:shrink-can-run-more-tests-count #f)   ;; DELAY TWEAKER (still needed?)
		  (if (and (null? tal)(null? reg))
		      (loop hed tal reg reruns)
		      (loop (runs:queue-next-hed tal reg reglen regfull)
			    (runs:queue-next-tal tal reg reglen regfull)
			    (let ((newl (append reg (list hed))))
			      (if regfull 
				  (cdr newl)
				  newl))
			    reruns)))
		 ;; At this point hed test registration must be completed.
		 ((eq? (hash-table-ref/default test-registry (runs:make-full-test-name test-name item-path) #f)
		       'start)
		  (debug:print-info 0 "Waiting on test registration(s): " (string-intersperse 
									   (filter (lambda (x)
										     (eq? (hash-table-ref/default test-registry x #f) 'start))
										   (hash-table-keys test-registry))
									   ", "))
		  (thread-sleep! 0.1)
		  (loop hed tal reg reruns))
		 ((not have-resources) ;; simply try again after waiting a second
		  (debug:print-info 1 "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! 1) ;; (+ 2 *global-delta*))
		  ;; could have done hed tal here but doing car/cdr of newtal to rotate tests
		  (loop (car newtal)(cdr newtal) reg reruns))
		 ((and have-resources
		       (or (null? prereqs-not-met)
			   (and (eq? testmode 'toplevel)
				(null? non-completed))))
		  (run:test run-id run-info keyvals runname test-record flags #f)
		  (hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'running)
		  (cdb:remote-run runs:shrink-can-run-more-tests-count #f)  ;; DELAY TWEAKER (still needed?)
		  ;; (thread-sleep! *global-delta*)
		  (if (not (null? tal))
		      (loop (runs:queue-next-hed tal reg reglen regfull)
			    (runs:queue-next-tal tal reg reglen regfull)
			    (runs:queue-next-reg tal reg reglen regfull)
			    reruns)))
		 (else ;; must be we have unmet prerequisites
		  (debug:print 4 "FAILS: " fails)
		  ;; If one or more of the prereqs-not-met are FAIL then we can issue
		  ;; a message and drop hed from the items to be processed.
		  (if (null? fails)
		      (begin
			;; couldn't run, take a breather
			(debug:print-info 4 "Shouldn't really get here, race condition? Unable to launch more tests at this moment, killing time ...")
			;; (thread-sleep! (+ 0.01 *global-delta*)) ;; long sleep here - no resources, may as well be patient
			;; we made new tal by sticking hed at the back of the list
			(loop (car newtal)(cdr newtal) reg reruns))
		      ;; the waiton is FAIL so no point in trying to run hed ever again
		      (if (not (null? tal))
			  (if (vector? hed)
			      (begin 
				(debug:print 1 "WARN: Dropping test " (db:test-get-testname hed) "/" (db:test-get-item-path hed)
					     " from the launch list as it has prerequistes that are FAIL")
				(cdb:remote-run runs:shrink-can-run-more-tests-count #f) ;; DELAY TWEAKER (still needed?)
				;; (thread-sleep! *global-delta*)
				(hash-table-set! test-registry (runs:make-full-test-name test-name item-path) 'removed)
				(loop (runs:queue-next-hed tal reg reglen regfull)
				      (runs:queue-next-tal tal reg reglen regfull)
				      (runs:queue-next-reg tal reg reglen regfull)
				      (cons hed reruns)))
			      (begin
				(debug:print 1 "WARN: Test not processed correctly. Could be a race condition in your test implementation? " hed) ;;  " as it has prerequistes that are FAIL. (NOTE: hed is not a vector)")
				(cdb:remote-run runs:shrink-can-run-more-tests-count #f) ;; DELAY TWEAKER (still needed?)
				;; (thread-sleep! (+ 0.01 *global-delta*))
				(loop hed tal reg reruns))))))))) ;; END OF INNER COND
	     
	     ;; case where an items came in as a list been processed
	     ((and (list? items)     ;; thus we know our items are already calculated
		   (not   itemdat)) ;; and not yet expanded into the list of things to be done
	      (if (and (debug:debug-mode 1) ;; (>= *verbosity* 1)
		       (> (length items) 0)
		       (> (length (car items)) 0))
		  (pp items))
	      (for-each
	       (lambda (my-itemdat)
		 (let* ((new-test-record (let ((newrec (make-tests:testqueue)))
					   (vector-copy! test-record newrec)
					   newrec))
			(my-item-path (item-list->path my-itemdat)))
		   (if (tests:match test-patts hed my-item-path required: required-tests) ;; (patt-list-match my-item-path item-patts)           ;; yes, we want to process this item, NOTE: Should not need this check here!
		       (let ((newtestname (runs:make-full-test-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)
			 (set! tal (cons newtestname tal)))))) ;; since these are itemized create new test names testname/itempath
	       items)
	      (if (not (null? tal))
		  (begin
		    (debug:print-info 4 "End of items list, looping with next after short delay")
		    ;; (thread-sleep! (+ 0.01 *global-delta*))
		    (loop (runs:queue-next-hed tal reg reglen regfull)
			  (runs:queue-next-tal tal reg reglen regfull)
			  (runs:queue-next-reg tal reg reglen regfull)
			  reruns))))

	     ;; if items is a proc then need to run items:get-items-from-config, get the list and loop 
	     ;;    - but only do that if resources exist to kick off the job
	     ((or (procedure? items)(eq? items 'have-procedure))
	      (let ((can-run-more    (cdb:remote-run runs:can-run-more-tests #f jobgroup max-concurrent-jobs)))
		(if (and (list? can-run-more)
			 (car can-run-more))
		    (let* ((prereqs-not-met (mt:get-prereqs-not-met run-id waitons item-path mode: testmode))
			   (fails           (runs:calc-fails prereqs-not-met))
			   (non-completed   (runs:calc-not-completed prereqs-not-met)))
		      (debug:print-info 8 "can-run-more: " can-run-more
					"\n testname:        " hed
					"\n prereqs-not-met: " (runs:pretty-string prereqs-not-met)
					"\n non-completed:   " (runs:pretty-string non-completed) 
					"\n fails:           " (runs:pretty-string fails)
					"\n testmode:        " testmode
					"\n num-retries:     " num-retries
					"\n (eq? testmode 'toplevel): " (eq? testmode 'toplevel)
					"\n (null? non-completed):    " (null? non-completed)
					"\n reruns:          " reruns
					"\n items:           " items
					"\n can-run-more:    " can-run-more)
		      ;; (thread-sleep! (+ 0.01 *global-delta*))
		      (cond ;; INNER COND #2
		       ((or (null? prereqs-not-met) ;; all prereqs met, fire off the test
			    ;; or, if it is a 'toplevel test and all prereqs not met are COMPLETED then launch
			    (and (eq? testmode 'toplevel)
				 (null? non-completed)))
			(let ((test-name (tests:testqueue-get-testname test-record)))
			  (setenv "MT_TEST_NAME" test-name) ;; 
			  (setenv "MT_RUNNAME"   runname)
			  (set-megatest-env-vars run-id inrunname: runname) ;; these may be needed by the launching process
			  (let ((items-list (items:get-items-from-config tconfig)))
			    (if (list? items-list)
				(begin
				  (tests:testqueue-set-items! test-record items-list)
				  ;; (thread-sleep! *global-delta*)
				  (loop hed tal reg reruns))
				(begin
				  (debug:print 0 "ERROR: The proc from reading the setup did not yield a list - please report this")
				  (exit 1))))))
		       ((null? fails)
			(debug:print-info 4 "fails is null, moving on in the queue but keeping " hed " for now")
			;; only increment num-retries when there are no tests runing
			(if (eq? 0 (list-ref can-run-more 1))
			    (begin
			      ;; TRY (if (> num-retries 100) ;; first 100 retries are low time cost
			      ;; TRY     (thread-sleep! (+ 2 *global-delta*))
			      ;; TRY     (thread-sleep! (+ 0.01 *global-delta*)))
			      (set! num-retries (+ num-retries 1))))
			(if (> num-retries  max-retries)
			    (if (not (null? tal))
				(loop (runs:queue-next-hed tal reg reglen regfull)
				      (runs:queue-next-tal tal reg reglen regfull)
				      (runs:queue-next-reg tal reg reglen regfull)
				      reruns))
			    (loop (car newtal)(cdr newtal) reg reruns))) ;; an issue with prereqs not yet met?
		       ((and (not (null? fails))(eq? testmode 'normal))
			(debug:print-info 1 "test "  hed " (mode=" testmode ") has failed prerequisite(s); "
				     (string-intersperse (map (lambda (t)(conc (db:test-get-testname t) ":" (db:test-get-state t)"/"(db:test-get-status t))) fails) ", ")
				     ", removing it from to-do list")
			(if (not (null? tal))
			    (begin
			      ;; (thread-sleep! *global-delta*)
			      (loop (runs:queue-next-hed tal reg reglen regfull)
				    (runs:queue-next-tal tal reg reglen regfull)
				    (runs:queue-next-reg tal reg reglen regfull)
				    (cons hed reruns)))))
		       (else
			(debug:print 8 "ERROR: No handler for this condition.")
			;; TRY (thread-sleep! (+ 1 *global-delta*))
			(loop (car newtal)(cdr newtal) reg reruns)))) ;; END OF IF CAN RUN MORE

		    ;; if can't run more just loop with next possible test
		    (begin
		      (debug:print-info 4 "processing the case with a lambda for items or 'have-procedure. Moving through the queue without dropping " hed)
		      ;; (thread-sleep! (+ 2 *global-delta*))
		      (loop (car newtal)(cdr newtal) reg reruns))))) ;; END OF (or (procedure? items)(eq? items 'have-procedure))
	     
	     ;; this case should not happen, added to help catch any bugs
	     ((and (list? items) itemdat)
	      (debug:print 0 "ERROR: Should not have a list of items in a test and the itemspath set - please report this")
	      (exit 1))
	     ((not (null? reruns))
	      (let* ((newlst (tests:filter-non-runnable run-id tal test-records)) ;; i.e. not FAIL, WAIVED, INCOMPLETE, PASS, KILLED,
		     (junked (lset-difference equal? tal newlst)))
		(debug:print-info 4 "full drop through, if reruns is less than 100 we will force retry them, reruns=" reruns ", tal=" tal)
		(if (< num-retries max-retries)
		    (set! newlst (append reruns newlst)))
		(set! num-retries (+ num-retries 1))
		;; (thread-sleep! (+ 1 *global-delta*))
		(if (not (null? newlst))
		    ;; since reruns have been tacked on to newlst create new reruns from junked
		    (loop (car newlst)(cdr newlst) reg (delete-duplicates junked)))))
	     ((not (null? tal))
	      (debug:print-info 4 "I'm pretty sure I shouldn't get here."))
	     ((not (null? reg)) ;; could we get here with leftovers?
	      (debug:print-info 0 "Have leftovers!")
	      (loop (car reg)(cdr reg) '() reruns))
	     (else
	      (debug:print-info 4 "Exiting loop with...\n  hed=" hed "\n  tal=" tal "\n  reruns=" reruns))
	     )))) ;; LET* ((test-record

    ;; we get here on "drop through" - loop for next test in queue
    ;; FIXME!!!! THIS SHOULD NOT REQUIRE AN EXIT!!!!!!!
    
    (debug:print-info 1 "All tests launched")
    (thread-sleep! 0.5)
    ;; FIXME! This harsh exit should not be necessary....
    ;; (if (not *runremote*)(exit)) ;; 
    #f)) ;; return a #f as a hint that we are done
;; Here we need to check that all the tests remaining to be run are eligible to run
;; and are not blocked by failed

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

;; runs:get-runs-by-patt
;; get runs by list of criteria
;; register a test run with the db
;;
;; Use: (db-get-value-by-header (db:get-header runinfo)(db:get-row runinfo))
;;  to extract info from the structure returned
;;
(define (runs:get-runs-by-patt db keys runnamepatt targpatt) ;; test-name)
  (let* ((tmp      (runs:get-std-run-fields keys '("id" "runname" "state" "status" "owner" "event_time")))
	 (keystr   (car tmp))
	 (header   (cadr tmp))
	 (res     '())
	 (key-patt "")
	 (runwildtype (if (substring-index "%" runnamepatt) "like" "glob"))
	 (qry-str  #f)
	 (keyvals  (keys:target->keyval keys targpatt)))
    (for-each (lambda (keyval)
		(let* ((key    (car keyval))
		       (patt   (cadr keyval))
		       (fulkey (conc ":" key))
		       (wildtype (if (substring-index "%" patt) "like" "glob")))
		  (if patt
		      (set! key-patt (conc key-patt " AND " key " " wildtype " '" patt "'"))
		      (begin
			(debug:print 0 "ERROR: searching for runs with no pattern set for " fulkey)
			(exit 6)))))
	      keyvals)
    (set! qry-str (conc "SELECT " keystr " FROM runs WHERE runname " runwildtype " ? " key-patt ";"))
    (debug:print-info 4 "runs:get-runs-by-patt qry=" qry-str " " runnamepatt)
    (sqlite3:for-each-row 
     (lambda (a . r)
       (set! res (cons (list->vector (cons a r)) res)))
     db 
     qry-str
     runnamepatt)
    (vector header res)))

(define (runs:test-get-full-path test)
  (let* ((testname (db:test-get-testname   test))
	 (itempath (db:test-get-item-path test)))
    (conc testname (if (equal? itempath "") "" (conc "(" itempath ")")))))

;; This is the *new* methodology. One record to inform them and in the chaos, organise them.
;;

		(for-each (lambda (varval)
			    (set! envdat (append envdat (list varval)))
			    (setenv (car varval)(cadr varval)))
			  (configf:get-section runconfig section)))
	      (list "default" target))
    (vector target runname testpatt keys keyvals envdat mconfig runconfig serverdat transport db toppath run-id)))

	 
(define (set-megatest-env-vars run-id #!key (inkeys #f)(inrunname #f)(inkeyvals #f))
  (let* ((target      (or (args:get-arg "-reqtarg")
			  (args:get-arg "-target")
			  (get-environment-variable "MT_TARGET")))
	 (keys    (if inkeys    inkeys    (cdb:remote-run db:get-keys #f)))
	 (keyvals (if inkeyvals inkeyvals (keys:target->keyval keys target)))
	 (vals (hash-table-ref/default *env-vars-by-run-id* run-id #f)))

;; Every time can-run-more-tests is called increment the delay
;;
;; NOTE: We run this server-side!! Do not use this global except in the runs:can-run-more-tests routine
;;
(define *last-num-running-tests* 0)
(define *runs:can-run-more-tests-count* 0)
(define (runs:shrink-can-run-more-tests-count db) ;; the db is just so we can use cdb:remote-run
  (set! *runs:can-run-more-tests-count* 0)) ;; (/ *runs:can-run-more-tests-count* 2)))

(define (runs:can-run-more-tests db jobgroup max-concurrent-jobs)
  (thread-sleep! (cond
		  ((> *runs:can-run-more-tests-count* 20) 2);; obviously haven't had any work to do for a while
		  (else 0)))
  (let* ((num-running             (db:get-count-tests-running db))
	 (num-running-in-jobgroup (db:get-count-tests-running-in-jobgroup db jobgroup))
	 (job-group-limit         (config-lookup *configdat* "jobgroups" jobgroup)))
    (if (> (+ num-running num-running-in-jobgroup) 0)
	(set! *runs:can-run-more-tests-count* (+ *runs:can-run-more-tests-count* 1)))
    (if (not (eq? *last-num-running-tests* num-running))
	(begin
	  (debug:print 2 "max-concurrent-jobs: " max-concurrent-jobs ", num-running: " num-running)
	  (set! *last-num-running-tests* num-running)))

				       (>= num-running-in-jobgroup job-group-limit))
				  (debug:print 1 "WARNING: number of jobs " num-running-in-jobgroup 
					       " in " jobgroup " exceeded, will not run " (tests:testqueue-get-testname test-record))
				  #t)
				 (else #f))))
	  (list (not can-not-run-more) num-running num-running-in-jobgroup max-concurrent-jobs job-group-limit)))))

;;======================================================================
;; New methodology. These routines will replace the above in time. For
;; now the code is duplicated. This stuff is initially used in the monitor
;; based code.
;;======================================================================


;; This is a duplicate of run-tests (which has been deprecated). Use this one instead of run tests.
;; keyvals.
;;
;;  test-names: Comma separated patterns same as test-patts but used in selection 
;;              of tests to run. The item portions are not respected.
;;              FIXME: error out if /patt specified
;;            
(define (runs:run-tests target runname test-patts user flags) ;; test-names
  (common:clear-caches) ;; clear all caches
  (let* ((db          #f)
	 (keys        (keys:config-get-fields *configdat*))
	 (keyvals     (keys:target->keyval keys target))
	 (run-id      (cdb:remote-run db:register-run #f keyvals runname "new" "n/a" user))  ;;  test-name)))
	 (deferred    '()) ;; delay running these since they have a waiton clause
	 ;; keepgoing is the defacto modality now, will add hit-n-run a bit later
	 ;; (keepgoing   (hash-table-ref/default flags "-keepgoing" #f))
	 (runconfigf   (conc  *toppath* "/runconfigs.config"))
	 (required-tests '())
	 (test-records (make-hash-table))
	 (all-test-names (tests:get-valid-tests *toppath* "%"))) ;; we need a list of all valid tests to check waiton names

    (set-megatest-env-vars run-id inkeys: keys) ;; these may be needed by the launching process

    (if (file-exists? runconfigf)
	(setup-env-defaults runconfigf run-id *already-seen-runconfig-info* keyvals "pre-launch-env-vars")
	(debug:print 0 "WARNING: You do not have a run config file: " runconfigf))
    
    ;; look up all tests matching the comma separated list of globs in
    ;; test-patts (using % as wildcard)

    (set! test-names (tests:get-valid-tests *toppath* test-patts))
    (set! test-names (delete-duplicates test-names))

    (debug:print-info 0 "test names " test-names)

    ;; on the first pass or call to run-tests set FAILS to NOT_STARTED if
    ;; -keepgoing is specified
    (if (eq? *passnum* 0)
	(begin
	  ;; have to delete test records where NOT_STARTED since they can cause -keepgoing to 
	  ;; get stuck due to becoming inaccessible from a failed test. I.e. if test B depends 
	  ;; on test A but test B reached the point on being registered as NOT_STARTED and test
	  ;; A failed for some reason then on re-run using -keepgoing the run can never complete.
	  (cdb:delete-tests-in-state *runremote* run-id "NOT_STARTED")
	  (cdb:remote-run db:set-tests-state-status #f run-id test-names #f "FAIL" "NOT_STARTED" "FAIL")))

    ;; from here on out the db will be opened and closed on every call runs:run-tests-queue
    ;; (sqlite3:finalize! db) 
    ;; now add non-directly referenced dependencies (i.e. waiton)



    (if (not (null? test-names))
	(let loop ((hed (car test-names))
		   (tal (cdr test-names)))         ;; 'return-procs tells the config reader to prep running system but return a proc
	  (let* ((config  (tests:get-testconfig hed 'return-procs))
		 (waitons (let ((instr (if config 
					   (config-lookup config "requirements" "waiton")
					   (begin ;; No config means this is a non-existant test
					     (debug:print 0 "ERROR: non-existent required test \"" hed "\"")
					     (if db (sqlite3:finalize! db))
					     (exit 1)))))
			    (debug:print-info 8 "waitons string is " instr)
			    (let ((newwaitons
				   (string-split (cond
						  ((procedure? instr)
						   (let ((res (instr)))
						     (debug:print-info 8 "waiton procedure results in string " res " for test " hed)

		  (loop (car remtests)(cdr remtests)))))))

    (if (not (null? required-tests))
	(debug:print-info 1 "Adding " required-tests " to the run queue"))
    ;; NOTE: these are all parent tests, items are not expanded yet.
    (debug:print-info 4 "test-records=" (hash-table->alist test-records))
    (let ((reglen (configf:lookup *configdat* "setup" "runqueue")))
      (if (equal? reglen "classic")
	  (runs:run-tests-queue-classic run-id runname test-records keyvals flags test-patts required-tests)
	  (runs:run-tests-queue-new     run-id runname test-records keyvals flags test-patts required-tests (any->number reglen))))

    (debug:print-info 4 "All done by here")))




















































































































































































































































































































































































































(define (runs:calc-fails prereqs-not-met)
  (filter (lambda (test)
	    (and (vector? test) ;; not (string? test))
		 (equal? (db:test-get-state test) "COMPLETED")
		 (not (member (db:test-get-status test)
			      '("PASS" "WARN" "CHECK" "WAIVED" "SKIP")))))
	  prereqs-not-met))

	     (conc t)
	     (conc (db:test-get-testname t) ":" (db:test-get-state t) "/" (db:test-get-status t))))
       lst))

(define (runs:make-full-test-name testname itempath)
  (if (equal? itempath "") testname (conc testname "/" itempath)))

(define (runs:queue-next-hed tal reg n regful)
  (if regful
      (if (null? reg) ;; doesn't make sense, this is probably NOT the problem of the car
	  (car tal)
	  (car reg))
      (car tal)))

(define (runs:queue-next-tal tal reg n regful)
  (if regful
      tal
      (let ((newtal (cdr tal)))
	(if (null? newtal)
	    reg
	    newtal
	    ))))

(define (runs:queue-next-reg tal reg n regful)
  (if regful
      (cdr reg)
      (if (eq? (length tal) 1)
	  '()
	  reg)))

(include "run-tests-queue-classic.scm")
(include "run-tests-queue-new.scm")

;; parent-test is there as a placeholder for when parent-tests can be run as a setup step
(define (run:test run-id run-info keyvals runname test-record flags parent-test)
  ;; All these vars might be referenced by the testconfig file reader
  (let* ((test-name    (tests:testqueue-get-testname   test-record))
	 (test-waitons (tests:testqueue-get-waitons    test-record))
	 (test-conf    (tests:testqueue-get-testconfig test-record))
	 (itemdat      (tests:testqueue-get-itemdat    test-record))
	 (test-path    (conc *toppath* "/tests/" test-name)) ;; could use tests:get-testconfig here ...
	 (force        (hash-table-ref/default flags "-force" #f))
	 (rerun        (hash-table-ref/default flags "-rerun" #f))
	 (keepgoing    (hash-table-ref/default flags "-keepgoing" #f))
	 (item-path     "")
	 (db           #f))
    (debug:print 4
		 "test-config: " (hash-table->alist test-conf)
		 "\n   itemdat: " itemdat
		 )
    ;; setting itemdat to a list if it is #f
    (if (not itemdat)(set! itemdat '()))
    (set! item-path (item-list->path itemdat))






    (debug:print 2 "Attempting to launch test " test-name (if (equal? item-path "/") "/" item-path))
    (setenv "MT_TEST_NAME" test-name) ;; 
    (setenv "MT_RUNNAME"   runname)
    (set-megatest-env-vars run-id inrunname: runname) ;; these may be needed by the launching process
    (change-directory *toppath*)

    ;; Here is where the test_meta table is best updated
    ;; Yes, another use of a global for caching. Need a better way?
    (if (not (hash-table-ref/default *test-meta-updated* test-name #f))
        (begin
	   (hash-table-set! *test-meta-updated* test-name #t)
           (runs:update-test_meta test-name test-conf)))
    
    ;; (lambda (itemdat) ;;; ((ripeness "overripe") (temperature "cool") (season "summer"))
    (let* ((new-test-path (string-intersperse (cons test-path (map cadr itemdat)) "/"))
	   (new-test-name (if (equal? item-path "") test-name (conc test-name "/" item-path))) ;; just need it to be unique
	   (test-id       (cdb:remote-run db:get-test-id #f  run-id test-name item-path))
	   (testdat       (cdb:get-test-info-by-id *runremote* test-id)))
      (if (not testdat)
	  (begin
	    ;; ensure that the path exists before registering the test
	    ;; NOPE: Cannot! Don't know yet which disk area will be assigned....
	    ;; (system (conc "mkdir -p " new-test-path))

	    ;; not -rerun and PASS, WARN or CHECK, do no run
	    ((and (or (not rerun)
		      keepgoing)
		  ;; Require to force re-run for COMPLETED or *anything* + PASS,WARN or CHECK
		  (or (member (test:get-status testdat) '("PASS" "WARN" "CHECK" "SKIP"))
		      (member (test:get-state  testdat) '("COMPLETED")))) 
	     (debug:print-info 2 "running test " test-name "/" item-path " suppressed as it is " (test:get-state testdat) " and " (test:get-status testdat))

	     (set! runflag #f))
	    ;; -rerun and status is one of the specifed, run it
	    ((and rerun
		  (let* ((rerunlst   (string-split rerun ","))
			 (must-rerun (member (test:get-status testdat) rerunlst)))
		    (debug:print-info 3 "-rerun list: " rerun ", test-status: " (test:get-status testdat)", must-rerun: " must-rerun)
		    must-rerun))
	     (debug:print-info 2 "Rerun forced for test " test-name "/" item-path)
	     (set! runflag #t))
	    ;; -keepgoing, do not rerun FAIL
	    ((and keepgoing
		  (member (test:get-status testdat) '("FAIL")))
	     (set! runflag #f))
	    ((and (not rerun)
		  (member (test:get-status testdat) '("FAIL" "n/a")))
	     (set! runflag #t))
	    (else (set! runflag #f)))
	   (debug:print 6 "RUNNING => runflag: " runflag " STATE: " (test:get-state testdat) " STATUS: " (test:get-status testdat))
	   (if (not runflag)
	       (if (not parent-test)
		   (debug:print 1 "NOTE: Not starting test " new-test-name " as it is state \"" (test:get-state testdat) 
				"\" and status \"" (test:get-status testdat) "\", use -rerun \"" (test:get-status testdat)
                                "\" or -force to override"))
	       ;; NOTE: No longer be checking prerequisites here! Will never get here unless prereqs are
	       ;;       already met.
	       ;; This would be a great place to do the process-fork
	       (if (not (launch-test test-id run-id run-info keyvals runname test-conf test-name test-path itemdat flags))
		   (begin
		     (print "ERROR: Failed to launch the test. Exiting as soon as possible")
		     (set! *globalexitstatus* 1) ;; 
		     (process-signal (current-process-id) signal/kill))))))
	((KILLED) 
	 (debug:print 1 "NOTE: " new-test-name " is already running or was explictly killed, use -force to launch it."))
	((LAUNCHED REMOTEHOSTSTART RUNNING)  
	 (if (> (- (current-seconds)(+ (db:test-get-event_time testdat)
				       (db:test-get-run_duration testdat)))
		600) ;; i.e. no update for more than 600 seconds
	     (begin
	       (debug:print 0 "WARNING: Test " test-name " appears to be dead. Forcing it to state INCOMPLETE and status STUCK/DEAD")
	       (tests:test-set-status! test-id "INCOMPLETE" "STUCK/DEAD" "Test is stuck or dead" #f))
	     (debug:print 2 "NOTE: " test-name " is already running")))
	(else       (debug:print 0 "ERROR: Failed to launch test " new-test-name ". Unrecognised state " (test:get-state testdat)))))))

;;======================================================================
;; END OF NEW STUFF
;;======================================================================

(define (get-dir-up-n dir . params) 
  (let ((dparts  (string-split dir "/"))

;;
;; NB// should pass in keys?
;;
(define (runs:operate-on action target runnamepatt testpatt #!key (state #f)(status #f)(new-state-status #f))
  (common:clear-caches) ;; clear all caches
  (let* ((db           #f)
	 (keys         (cdb:remote-run db:get-keys db))
	 (rundat       (cdb:remote-run runs:get-runs-by-patt db keys runnamepatt target))
	 (header       (vector-ref rundat 0))
	 (runs         (vector-ref rundat 1))
	 (states       (if state  (string-split state  ",") '()))
	 (statuses     (if status (string-split status ",") '()))
	 (state-status (if (string? new-state-status) (string-split new-state-status ",") '(#f #f))))
    (debug:print-info 4 "runs:operate-on => Header: " header " action: " action " new-state-status: " new-state-status)
    (if (> 2 (length state-status))

;;======================================================================
;; Lock/unlock runs
;;======================================================================

(define (runs:handle-locking target keys runname lock unlock user)
  (let* ((db       #f)
	 (rundat   (cdb:remote-run runs:get-runs-by-patt db keys runname target))
	 (header   (vector-ref rundat 0))
	 (runs     (vector-ref rundat 1)))
    (for-each (lambda (run)
		(let ((run-id (db:get-value-by-header run header "id")))
		  (if (or lock
			  (and unlock
			       (begin

	     (begin
	       (print "Updating " test-name " " fld " to " val)
	       (cdb:remote-run db:testmeta-update-field #f test-name fld val)))))
     '(("author" 2)("owner" 3)("description" 4)("reviewed" 5)("tags" 9)))))

;; Update test_meta for all tests
(define (runs:update-all-test_meta db)
  (let ((test-names (get-all-legal-tests)))
    (for-each 
     (lambda (test-name)
       (let* ((test-path    (conc *toppath* "/tests/" test-name))
	      (test-configf (conc test-path "/testconfig"))
	      (testexists   (and (file-exists? test-configf)(file-read-access? test-configf)))
	      ;; read configs with tricks turned off (i.e. no system)
	      (test-conf    (if testexists (read-config test-configf #f #f)(make-hash-table))))

;;======================================================================
;; Copyright 2006-2012, Matthew Welland.
;; 
;;  This program is made available under the GNU GPL version 2.0 or
;;  greater. See the accompanying file COPYING for details.
;; 
;;  This program is distributed WITHOUT ANY WARRANTY; without even the
;;  implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
;;  PURPOSE.

		      (release-dot-lock outputfilename)))
		  (close-output-port oup)
		  (change-directory orig-dir)
		  ;; NB// tests:test-set-toplog! is remote internal...
		  (tests:test-set-toplog! db run-id test-name outputfilename)
		  )))))))




(define (get-all-legal-tests)

  (let* ((tests  (glob (conc *toppath* "/tests/*")))


	 (res    '()))
    (debug:print-info 4 "Looking at tests " (string-intersperse tests ","))

    (for-each (lambda (testpath)
		(if (file-exists? (conc testpath "/testconfig"))
		    (set! res (cons (last (string-split testpath "/")) res))))
	      tests)
    res))

(define (tests:get-testconfig test-name system-allowed)
  (let* ((test-path    (conc *toppath* "/tests/" test-name))
	 (test-configf (conc test-path "/testconfig"))
	 (testexists   (and (file-exists? test-configf)(file-read-access? test-configf))))
    (if testexists
	(read-config test-configf #f system-allowed environ-patt: (if system-allowed

					 (member (db:test-get-state wtdat)
						 '("INCOMPETE")))
				     (set! keep-test #f)))) ;; no point in running this one again
			     waitons))))
	 (if keep-test (set! runnables (cons testkeyname runnables)))))
     testkeynames)
    runnables))






















































































;;======================================================================
;; test steps
;;======================================================================

;; teststep-set-status! used to be here

# run some tests

BINPATH=$(PWD)/../bin
MEGATEST=$(BINPATH)/megatest
PATH := $(BINPATH):$(PATH)
RUNNAME := $(shell date +w%V.%u.%H.%M)
IPADDR := "-"
# Set SERVER to "-server -"
SERVER  = 
DEBUG   = 1

	@echo "WARNING: No longer running fullprep, test converage may be lessened"
	cd fullrun;sleep 0;$(MEGATEST) -runtests % -target $(TARGET) :runname $(RUNNAME)_aa -debug $(DEBUG) $(LOGGING) > aa.log 2> aa.log &
	cd fullrun;sleep 0;$(MEGATEST) -runtests % -target $(TARGET) :runname $(RUNNAME)_ab -debug $(DEBUG) $(LOGGING) > ab.log 2> ab.log &
	cd fullrun;sleep 0;$(MEGATEST) -runtests % -target $(TARGET) :runname $(RUNNAME)_ac -debug $(DEBUG) $(LOGGING) > ac.log 2> ac.log &
	cd fullrun;sleep 0;$(MEGATEST) -runtests % -target $(TARGET) :runname $(RUNNAME)_ad -debug $(DEBUG) $(LOGGING) > ad.log 2> ad.log &	
#	cd fullrun;sleep 0;$(MEGATEST) -runtests % -target $(TARGET) :runname $(RUNNAME)_ae -debug $(DEBUG) $(LOGGING) > ae.log 2> ae.log &	
#	cd fullrun;sleep 0;$(MEGATEST) -runtests % -target $(TARGET) :runname $(RUNNAME)_af -debug $(DEBUG) $(LOGGING) > af.log 2> af.log &	
	cd fullrun;sleep 10;$(MEGATEST) -run-wait  -target $(TARGET) :runname % -testpatt % :state RUNNING,LAUNCHED;echo ALL DONE

test6: fullprep
	cd fullrun;$(MEGATEST) -runtests runfirst -testpatt %/1 -reqtarg ubuntu/nfs/none :runname $(RUNNAME)_itempatt -v
	cd fullrun;$(MEGATEST) -runtests runfirst -testpatt %blahha% -reqtarg ubuntu/nfs/none :runname $(RUNNAME)_itempatt -debug 10
	cd fullrun;$(MEGATEST) -rollup :runname newrun -target ubuntu/nfs/none -debug 10

test7: 
	@echo Only a/c testname c should remain. If there is a run a/b/c then there is a cache issue.
	(cd simplerun; \
	 $(MEGATEST) -server - -daemonize; \
         $(MEGATEST) -remove-runs -target %/% :runname % -testpatt %; \
         $(MEGATEST) -runtests %  -target a/b :runname c; sleep 5; \
	 $(MEGATEST) -remove-runs -target a/c :runname c; \
	 $(MEGATEST) -runtests %  -target a/c :runname c; \
	 $(MEGATEST) -remove-runs -target a/b :runname c -testpatt % ; \
	 $(MEGATEST) -runtests %  -target a/d :runname c;$(MEGATEST) -list-runs %|egrep ^Run:) > test7.log 2> test7.log 
	logpro test7.logpro test7.html < test7.log
	@echo 
	@echo Run \"firefox test7.html\" to see the results.



































cleanprep : ../*.scm Makefile */*.config
	mkdir -p fullrun/tmp/mt_runs fullrun/tmp/mt_links
	cd ..;make;make install
	rm -f */logging.db
	touch cleanprep

fullprep : cleanprep

	rm -f */megatest.db */logging.db */monitor.db || true
	killall -v mtest dboard || true

hardkill : kill
	sleep 5;killall -v mtest main.sh dboard -9

listservers :
	cd fullrun;$(MEGATEST) -listservers

runforever :
	while(ls); do runname=`date +%F-%R:%S`;(cd fullrun;$(MEGATEST) -runall -target ubuntu/nfs/none :runname $$runname;/home/matt/data/megatest/megatest -runall -target ubuntu/nfs/none :runname $$runname;/home/matt/data/megatest/megatest -runall -target ubuntu/nfs/none :runname $$runname);done

# Set launchwait to yes to use the old launch run code that waits for the launch process to return before 
# proceeding.
# launchwait yes

# If set to "default" the old code is used. Otherwise defaults to 200 or uses
# numeric value given.
#
# runqueue 2

# It is possible (but not recommended) to override the rsync command used
# to populate the test directories. For test development the following 
# example can be useful
#
# testcopycmd cp --remove-destination -rsv TEST_SRC_PATH/. TEST_TARG_PATH/. >> TEST_TARG_PATH/mt_launch.log 2>> TEST_TARG_PATH/mt_launch.log

				(string? (getenv "MT_RUN_AREA_HOME"))))

(test "server-register, get-best-server" #t (let ((res #f))
					      (open-run-close tasks:server-register tasks:open-db 1 "bob" 1234 100 'live 'http)
					      (set! res (open-run-close tasks:get-best-server tasks:open-db))
					      (number? (vector-ref res 3))))

(test "de-register server" #f (let ((res #f))
				(open-run-close tasks:server-deregister tasks:open-db "bob" port: 1234)
				(open-run-close tasks:get-best-server tasks:open-db)))

(define server-pid #f)
(test "launch server" #t (let ((pid (process-fork (lambda ()
						    ;; (daemon:ize)
						    (server:launch 'http)))))
			   (set! server-pid pid)
			   (number? pid)))