Megatest

    (11 "PREQ_DISCARDED")
    (12 "ABORT")))

(define *common:ended-states*       ;; states which indicate the test is stopped and will not proceed
  '("COMPLETED" "ARCHIVED" "KILLED" "KILLREQ" "STUCK" "INCOMPLETE" ))

(define *common:badly-ended-states* ;; these roll up as CHECK, i.e. results need to be checked
  '("KILLED" "KILLREQ" "STUCK" "INCOMPLETE" "DEAD" "CHECK"))

(define *common:well-ended-states* ;; an item's prereq in this state allows item to proceed
  '("PASS" "WARN" "WAIVED" "SKIP"))

;; BBnote: *common:running-states* used from db:set-state-status-and-roll-up-items
(define *common:running-states*     ;; test is either running or can be run
  '("RUNNING" "REMOTEHOSTSTART" "LAUNCHED" "STARTED"))

(define *common:cant-run-states*    ;; These are stopping conditions that prevent a test from being run
  '("COMPLETED" "KILLED" "UNKNOWN" "INCOMPLETE" "ARCHIVED"))

;;   count     - count down to zero, at some point we'd give up if the load never drops
;;   num-tries - count down to zero number tries to get numcpus
;;
(define (common:wait-for-cpuload maxnormload numcpus-in
				 #!key (count 1000)
				 (msg #f)(remote-host #f)(num-tries 5))
  (let* ((loadavg     (common:get-cpu-load remote-host))
    ;; not possible to have zero.  If we get 1, it's possible that we got the previous default, and we should check again
    (numcpus     (if (<= 1 numcpus-in)
    (common:get-num-cpus remote-host) numcpus-in))

    (first       (car loadavg))
    (next        (cadr loadavg))
    (adjmaxload  (* maxnormload (max 1 numcpus))) ;; possible bug where numcpus (or could be maxload) is zero, crude




						  ;; fallback is to at least use 1

    ;; effective load accounts for load jumps, this should elminate all the first-next-avg, adjwait, load-jump-limit
    ;; etc.
    (effective-load    (common:get-intercept first next))
    (recommended-delay (common:get-delay effective-load numcpus))
    (effective-host    (or remote-host "localhost"))
    (normalized-effective-load (/ effective-load numcpus))
    (will-wait                 (> normalized-effective-load maxnormload)))
    (if (and will-wait (> recommended-delay 1)) 
      (let* ((actual-delay (min recommended-delay 30)))
        (if (common:low-noise-print 30 (conc (round actual-delay) "-safe-load"))
          (debug:print-info 0 *default-log-port* "Load control, delaying "
				actual-delay " seconds to maintain safe load. current normalized effective load is "
				normalized-effective-load". maxnormload = " maxnormload " numcpus = " numcpus " loadavg = " loadavg " effective-load = " effective-load))
	(thread-sleep! actual-delay)))
    
    (cond
     ;; bad data, try again to get the data
     ((not will-wait)
      (if (common:low-noise-print 3600 (conc (round normalized-effective-load) "-load-acceptable-" effective-host))
      	  (debug:print 0 *default-log-port* "Effective load on " effective-host " is acceptable at " effective-load " continuing.")))

		(loop (car tal)(cdr tal) newr)))))))




;; the new prereqs calculation, looks also at itempath if specified
;; all prereqs must be met
;;    if prereq test with itempath='' is in common:well-ended-states, then prereq is met
;;    if prereq test with itempath=ref-item-path and COMPLETED with PASS, WARN, CHECK, or WAIVED then prereq is met
;;
;; Note: mode 'normal means that tests must be COMPLETED and ok (i.e. PASS, WARN, CHECK, SKIP or WAIVED)
;;       mode 'toplevel means that tests must be COMPLETED only
;;       mode 'itemmatch or 'itemwait means that tests items must be COMPLETED and (PASS|WARN|WAIVED|CHECK) [[ NB// NOT IMPLEMENTED YET ]]
;;       mode 'exclusive means this test/item cannot run if the same test/item is LAUNCHED,REMOTEHOSTSTART or RUNNING
;;
;; IDEA for consideration:
;;    1. collect all tests "upstream"
;;    2. any NOT completed and good? if yes => return those as prereqs not met, if no => return null list
;; 
;; (define (db:get-prereqs-not-met dbstruct run-id waitons ref-item-path mode)
(define (db:get-prereqs-not-met dbstruct run-id waitons ref-test-name ref-item-path mode itemmaps) ;; #!key (mode '(normal))(itemmap #f))
  ;; BBnote - rollup of an itemized test's overall state/status done in db:set-state-status-and-roll-up-items
  (debug:print 4 *default-log-port* "db:get-prereqs-not-met: " waitons)
  (append
   (if (member 'exclusive mode)
       (let ((running-tests (db:get-tests-for-run dbstruct
						  #f  ;; run-id of #f means for all runs. 
						  (if (string=? ref-item-path "")   ;; testpatt
						      ref-test-name
						      (conc ref-test-name "/" ref-item-path))

	;;	(if (equal? (db:test-get-item-path testdat) "")
	;;	    (db:test-get-testname testdat)
	;;	    (conc (db:test-get-testname testdat)
	;;		  "/"
	;;		  (db:test-get-item-path testdat))))
	 running-tests) ;; calling functions want the entire data
       '())



   ;; collection of: for each waiton -
   ;;   if this ref-test-name is an item in an itemized test and mode is itemwait/itemmatch:
   ;;     if waiton is not itemized - if waiton is not both completed and in ok status, add as unmet prerequisite
   ;;     if waiton is itemized:
   ;;           and waiton's items are not expanded, add as unmet prerequisite
   ;;           else if matching waiton item is not both completed and in an ok status, add as unmet prerequisite

                          (waiton-record   (hash-table-ref/default test-records waiton #f))
			  (waiton-tconfig  (if waiton-record (vector-ref waiton-record 1) #f))
			  (waiton-itemized (and waiton-tconfig
						(or (hash-table-ref/default waiton-tconfig "items" #f)
						    (hash-table-ref/default waiton-tconfig "itemstable" #f))))
			  (itemmaps        (tests:get-itemmaps config))  ;; (configf:lookup config "requirements" "itemmap"))
			  (new-test-patts  (tests:extend-test-patts test-patts hed waiton itemmaps hed-itemized-waiton))) 
		     (debug:print-info 2 *default-log-port* "Test " waiton " has " (if waiton-record "a" "no") " waiton-record and" (if waiton-itemized " " " no ") "items")
		     ;; need to account for test-patt here, if I am test "a", selected with a test-patt of "hed/b%"
		     ;; and we are waiting on "waiton" we need to add "waiton/,waiton/b%" to test-patt
		     ;; is this satisfied by merely appending "/" to the waiton name added to the list?
		     ;;
		     ;; This approach causes all of the items in an upstream test to be run 
		     ;; if we have this waiton already processed once we can analzye it for extending
		     ;; tests to be run, since we can't properly process waitons unless they have been

                                   (set! test-names (cons waiton test-names)) ;; need to process this one, only add once the waiton tconfig read
                                   (set! required-tests (cons (conc waiton "/") required-tests))
                                   (set! test-patts new-test-patts))
                                 (begin
                                   (debug:print-info 0 *default-log-port* "Waitor(s) not yet on testpatt for " waiton ", setting up to re-process it")
                                   (set! tal (append (cons waiton tal)(list hed)))))
                             (begin
                               (debug:print-info 2 *default-log-port* "Adding non-itemized test " waiton " to required-tests")
                               (set! required-tests (cons waiton required-tests))
                               (set! test-patts new-test-patts)))
			 (begin
			   (debug:print-info 0 *default-log-port* "No testconfig info yet for " waiton ", setting up to re-process it")
			   (set! tal (append (cons waiton tal)(list hed))))) ;; (cons (conc waiton "/") required-tests))
		     ;; NOPE: didn't work. required needs to be plain test names. Try tacking on to test-patts
		     ;;  - doesn't work

;; return value of runs:expand-items is passed back to runs-tests-queue and is fed to named loop with this signature:
;;    (let loop ((hed         (car sorted-test-names))
;;	         (tal         (cdr sorted-test-names))
;;	         (reg         '()) ;; registered, put these at the head of tal 
;;	         (reruns      '()))
(define (runs:expand-items hed tal reg reruns regfull newtal jobgroup max-concurrent-jobs run-id waitons item-path testmode test-record can-run-more items runname tconfig reglen test-registry test-records itemmaps)
  (let* ((loop-list       (list hed tal reg reruns))
         (junk (debug:print-info 4 *default-log-port* "expand-items calling rmt:get-prereqs-not-met"))
	 (prereqs-not-met (let ((res (rmt:get-prereqs-not-met run-id waitons hed item-path mode: testmode itemmaps: itemmaps)))
			    (if (list? res)
				res
				(begin
				  (debug:print 0 *default-log-port*
					       "ERROR: rmt:get-prereqs-not-met returned non-list!\n"
					       "  res=" res " run-id=" run-id " waitons=" waitons " hed=" hed " item-path=" item-path " testmode=" testmode " itemmaps=" itemmaps)

	   (list (car newtal)(append (cdr newtal) reg) '() reruns)
	  #f)) 
     ((null? runnables)
      (debug:print-info 4 *default-log-port* "cond branch - "  "ei-7")
      #f) ;; if we get here and non-completed is null then it is all over.
     (else
      (debug:print-info 4 *default-log-port* "cond branch - "  "ei-8")
      (debug:print 2 *default-log-port* "WARNING: FAILS or incomplete tests maybe preventing completion of this run. Watch for issues with test " hed ", continuing for now")
      (list (car newtal)(cdr newtal) reg reruns)))))

(define (runs:mixed-list-testname-and-testrec->list-of-strings inlst)
  (if (null? inlst)
      '()
      (map (lambda (t)
	     (cond

		    (not (member 'exclusive testmode)))))
      ;; (hash-table-delete! *max-tries-hash* (db:test-make-full-name test-name item-path))
      ;; we are going to reset all the counters for test retries by setting a new hash table
      ;; this means they will increment only when nothing can be run
      (set! *max-tries-hash* (make-hash-table))
      
      (run:test run-id run-info keyvals runname test-record flags #f test-registry all-tests-registry runsdat testdat)
      (if (debug:debug-mode 3)
        (runs:incremental-print-results run-id)
      )
      (hash-table-set! test-registry (db:test-make-full-name test-name item-path) 'running)
      (runs:shrink-can-run-more-tests-count runsdat)  ;; DELAY TWEAKER (still needed?)
      ;; (thread-sleep! *global-delta*)
      (if (or (not (null? tal))(not (null? reg)))
	  (runs:loop-values tal reg reglen regfull reruns) ;; hed should be dropped at this time
	  #f))
     

		  (debug:print-info 0 *default-log-port* "Skipping test " tfullname " as it has been marked do not run due to being completed or not runnable"))
	      (if (or (not (null? tal))(not (null? reg)))
		  (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 tal)(cdr tal) reg reruns))))
        (if (debug:debug-mode 3)
	  (runs:incremental-print-results run-id)
        )
	(debug:print 4 *default-log-port* "TOP OF LOOP => "
		     "test-name: " test-name
		     "\n  hed:         " hed
		     "\n  tal:         " (runs:pretty-long-list tal)
		     "\n  reg:         " reg
                     "\n  test-record  " test-record
                     "\n  itemdat:     " itemdat

		  (if loop-list
		      (apply loop loop-list)
                      (debug:print-info 4 *default-log-port* " -- Can't expand hed="hed)
                      )
                  )
		;; if can't run more just loop with next possible test
		(loop (car newtal)(cdr newtal) reg reruns))))
        

	 ;; this case should not happen, added to help catch any bugs
	 ((and (list? items) itemdat)
          (debug:print-info 4 *default-log-port* "cond branch - "  "rtq-5")
	  (debug:print-error 0 *default-log-port* "Should not have a list of items in a test and the itemspath set - please report this")
	  (exit 1))

	 ((not (null? reruns))
          (debug:print-info 4 *default-log-port* "cond branch - "  "rtq-6")
	  (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 *default-log-port* "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 *default-log-port* "cond branch - "  "rtq-7")
	  (debug:print-info 4 *default-log-port* "I'm pretty sure I shouldn't get here."))
	 ((not (null? reg)) ;; could we get here with leftovers?
          (debug:print-info 4 *default-log-port* "cond branch - "  "rtq-8")
	  (debug:print-info 0 *default-log-port* "Have leftovers!")
	  (loop (car reg)(cdr reg) '() reruns))

	     (set! runflag #t))

	    (else (set! runflag #f)))
	   (debug:print 4 *default-log-port* "RUNNING => runflag: " runflag " STATE: " (test:get-state testdat) " STATUS: " (test:get-status testdat))
	   (if (not runflag)
	       (if (not parent-test)
		   (if (runs:lownoise (conc "not starting test" full-test-name) 60)
		       (debug:print 3 *default-log-port* "NOTE: Not starting test " full-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
	       ;; 
	       (let ((skip-test   #f)

		    ;; seek and kill in flight -runtests with % as testpatt here
		    ;; (if (equal? testpatt "%")
		    (tasks:kill-runner target run-name testpatt)
		    ;; (debug:print 0 *default-log-port* "not attempting to kill any run launcher processes as testpatt is " testpatt))
		    (debug:print 1 *default-log-port* "Removing tests for run: " runkey " " (db:get-value-by-header run header "runname")))
		   ((set-state-status)
		    ;; (if (tasks:need-server run-id)(tasks:start-and-wait-for-server tdbdat run-id 10))
		    (debug:print 2 *default-log-port* "Modifying state and status for tests for run: " runkey " " (db:get-value-by-header run header "runname")))
		   ((print-run)
		    (debug:print 1 *default-log-port* "Printing info for run " runkey ", run=" run ", tests=" tests ", header=" header)
		    action)
		   ((run-wait)
		    (debug:print 1 *default-log-port* "Waiting for run " runkey ", run=" runnamepatt " to complete"))
		   ((archive)
		    (debug:print 1 *default-log-port* "Archiving/restoring (" (args:get-arg "-archive") ") data for run: " runkey " " (db:get-value-by-header run header "runname"))

				       )
				      ) ; end case rem-status
                                    ) ; end let
                                  ); end cond has-subrun

                                 (else
                                  ;; BB - TODO - consider backgrounding to threads to delete tests (work below) 
                                  (debug:print-info 2 *default-log-port* "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)

	;; 			      (thread-sleep! 1)
	 ;; 			      (loop (rmt:no-sync-get-lock lock-key) expire-time)))))))
	 )
    (case clean-mode
      ((remove-data-only)(mt:test-set-state-status-by-id (db:test-get-run_id test)(db:test-get-id test) "CLEANING" "LOCKED" #f))
      ((remove-all)      (mt:test-set-state-status-by-id (db:test-get-run_id test)(db:test-get-id test) "REMOVING" "LOCKED" #f))
      ((archive-remove)  (mt:test-set-state-status-by-id (db:test-get-run_id test)(db:test-get-id test) "ARCHIVE_REMOVING" #f #f)))
    (debug:print-info 2 *default-log-port* "Attempting to remove " (if real-dir (conc " dir " real-dir " and ") "") " link " run-dir)
    (if (and real-dir 
	     (> (string-length real-dir) 5)
	     (common:file-exists? real-dir)) ;; bad heuristic but should prevent /tmp /home etc.
	(let* ((realpath (resolve-pathname run-dir)))
	  (debug:print-info 1 *default-log-port* "Recursively removing " realpath)
	  (if (common:file-exists? realpath)
	      (runs:safe-delete-test-dir realpath)

	  (let* ((linktree (common:get-linktree)) ;; (if toppath (configf:lookup *configdat* "setup" "linktree")))
		 (runtop   (conc linktree "/" target "/" runname))
		 (files    (if (common:file-exists? runtop)
			       (append (glob (conc runtop "/.megatest*"))
				       (glob (conc runtop "/.runconfig*")))
			       '())))
	    (if (null? files)
		(debug:print-info 2 *default-log-port* "No cached megatest or runconfigs files found. None removed.")
		(begin
		  (debug:print-info 2 *default-log-port* "Removing cached files:\n    " (string-intersperse files "\n    "))
		  (for-each 
		   (lambda (f)
		     (handle-exceptions
			 exn
			 (debug:print 0 *default-log-port* "WARNING: Failed to remove file " f ", exn=" exn)
		       (delete-file f)))
		   files))))
	  (debug:print-error 0 *default-log-port* "-clean-cache requires -runname."))
      (debug:print-error 0 *default-log-port* "-clean-cache requires -target or -reqtarg")))

	     (not (string-match (conc "("curr-host "|" curr-host"\\..*)") target-host))
	     (not (equal? curr-ip target-host)))
	(begin
	  (debug:print-info 0 *default-log-port* "Starting server on " target-host ", logfile is " logfile)
	  (setenv "TARGETHOST" target-host)))
      
    (setenv "TARGETHOST_LOGF" logfile)
    (thread-sleep! (/ (random 3000) 1000)) ;; add a random initial delay. It seems pretty common that many running tests request a server at the same time
    (debug:print 0 *default-log-port* "INFO: starting server at " (common:human-time))
    (system (conc "nbfake " cmdln))
    (unsetenv "TARGETHOST_LOGF")
    (if (get-environment-variable "TARGETHOST")(unsetenv "TARGETHOST"))
    (thread-join! log-rotate)
    (pop-directory)))

(define (server:get-client-signature) ;; BB> why is this proc named "get-"?  it returns nothing -- set! has not return value.
  (if *my-client-signature* *my-client-signature*
      (let ((sig (server:mk-signature)))
        (set! *my-client-signature* sig)
        *my-client-signature*)))


;; if server-start-last exists, and wasn't old enough, wait <idle time> + 1, then call this function recursively until it is old enough.
;; if it is old enough, overwrite it and wait 0.25 seconds.
;; if it then has the wrong server key, wait <idle time> + 1 and call this function recursively.
;;
(define (server:wait-for-server-start-last-flag areapath)
  (let* ((start-flag (conc areapath "/logs/server-start-last"))
	 ;;; THIS INTERACTS WITH [server] timeout. Suggest using 0.1 or above for timeout (6 seconds)
	 (idletime    (configf:lookup-number *configdat* "server" "idletime" default: 4))
	 (server-key (conc (get-host-name) "-" (current-process-id))))
    (if (file-exists? start-flag)
	(let* ((fmodtime (file-modification-time start-flag))
	       (delta    (- (current-seconds) fmodtime))
	       (old-enough   (> delta idletime))
               (new-server-key "")
              )

          ;; write start-flag file, wait 0.25s, then if previously the start-flag file was older than <idletime> seconds, and the new file still has the same server key as you just wrote, return #t.
	  ;; the intention is to make sure nfs can read the file we just wrote, and make sure it was written by us, and not another process.
           (if (and old-enough
		   (begin
                     (debug:print-info 0 *default-log-port* "Writing " start-flag)
		     (with-output-to-file start-flag (lambda () (print server-key)))
		     (thread-sleep! 0.25)
		     (set! new-server-key (with-input-from-file start-flag (lambda () (read-line))))
		     (equal? server-key new-server-key))
                )
	      #t

           ;; If either of the above conditions is not true, print a "Gating server start" message, wait <idle-time> + 1, then call this function recursively. 
	      (begin
		(debug:print-info 0 *default-log-port* "Gating server start, last start: "
				  (seconds->time-string fmodtime) ", time since last start: " delta ", required idletime: " idletime ", gating reason:" (if old-enough "another job started a server" "too soon to start another server"))

		(thread-sleep! ( + 1 idletime))
		(server:wait-for-server-start-last-flag areapath)))))))


        
;; kind start up of server, wait before allowing another server for a given
;; area to be launched
;;
(define (server:kind-run areapath)
  ;; look for $MT_RUN_AREA_HOME/logs/server-start-last
  ;; and wait for it to be at least <server idletime> seconds old
  (server:wait-for-server-start-last-flag areapath)
  (if (not (server:check-if-running areapath)) ;; why try if there is already a server running?
      (let* (
	     (lock-file    (conc areapath "/logs/server-start.lock")))
	(let* ((start-flag (conc areapath "/logs/server-start-last")))
	  (common:simple-file-lock-and-wait lock-file expire-time: 25)
	  (debug:print-info  0 *default-log-port* "server:kind-run: touching " start-flag)
	  (system (conc "touch " start-flag)) ;; lazy but safe
	  (server:run areapath)
	  (thread-sleep! 20) ;; don't release the lock for at least a few seconds. And allow time for the server startup to get to "SERVER STARTED".
	  (common:simple-file-release-lock lock-file)))

      (debug:print-info 0 *default-log-port* "Found server already running. NOT trying to start another.")
   )
)

;; this one seems to be the general entry point

if [ "\$LD_LIBRARY_PATH" != "" ];then
  export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:\$LD_LIBRARY_PATH:$libdir
else
  export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$libdir
fi

export MT_SQLITE3_EXE=$sqlite3_exe

http_vars="http_proxy https_proxy HTTP_PROXY HTTPS_PROXY"
for i in \$http_vars
do
j=\${!i}
if [ "\$j" != "" ]; then
   unset \$i
fi
done
__EOF
) > $cfgfile
  echo 
else
  echo "INFO: LD_LIBRARY_PATH not set" >&2
fi