Megatest

              (set! ezsteps #t) ;; set the needed flag
	      (set! ezstepslst
                    (append (or ezstepslst '())
                            (list (list "subrun" (conc "{subrun=true} " mt-cmd)))))))

	;; process the ezsteps
	(if ezsteps
	    (let* ((envdbf        (conc "/tmp/."(current-user-name)"-"(current-process-id)"-"run-id"-"test-id".db"))
		   (all-steps-dat (make-hash-table))) ;; keep all the info around as stepname ==> alist;
	                                              ;;; where  'params is the params list (add other
	                                              ;;; stuff as needed)
	      (if (not (common:file-exists? ".ezsteps"))(create-directory ".ezsteps"))
	      ;; if ezsteps was defined then we are sure to have at least one step but check anyway
	      (if (not (> (length ezstepslst) 0))
		  (debug:print-error 0 *default-log-port* "ezsteps defined but ezstepslst is zero length")
                  (let ((all-step-names (map car ezstepslst))
                        (status-file (file-open "ezsteps.status" (+ open/append open/wronly open/creat)))
                        )
		   (setenv "MT_STEP_NAMES" (string-intersperse all-step-names " "))
		   (let loop ((ezstep (car ezstepslst))
			       (tal    (cdr ezstepslst))
			       (prevstep #f))
		      (debug:print-info 0 *default-log-port* "Processing ezstep \"" (string-intersperse ezstep " ") "\"")
		      ;; check exit-info (vector-ref exit-info 1)
		      (if (launch:einf-exit-status exit-info) ;; (vector-ref exit-info 1)
			  (let* ((logpro-used (launch:runstep ezstep run-id test-id exit-info m
							      tal testconfig all-steps-dat prevstep envdbf))
				 (stepname    (car ezstep))
				 (stepparms   (hash-table-ref all-steps-dat stepname)))
			    (setenv "MT_STEP_NAME" stepname)
			    (pp (hash-table->alist all-steps-dat))
			    ;; if logpro-used read in the stepname.dat file
			    (if (and logpro-used (common:file-exists? (conc stepname ".dat")))
				(launch:load-logpro-dat run-id test-id stepname))

                          (file-write status-file (conc stepname " " (launch:einf-exit-code exit-info) "\n"))

			    (if (steprun-good? logpro-used (launch:einf-exit-code exit-info) stepparms)
				(if (not (null? tal))
				    (loop (car tal) (cdr tal) stepname))
				(debug:print 0 *default-log-port* "WARNING: step " (car ezstep) " failed. Stopping")))
			(debug:print 0 *default-log-port* "WARNING: a prior step failed, stopping at " ezstep)))
                        (file-close status-file)
                        )

                        ))))))

(define (launch:monitor-job run-id test-id item-path fullrunscript ezsteps test-name tconfigreg exit-info m work-area runtlim misc-flags)
  (let* ((update-period (string->number (or (configf:lookup *configdat* "setup" "test-stats-update-period") "30")))
         (start-seconds (current-seconds))
	 (calc-minutes  (lambda ()
			  (inexact->exact 
			   (round 
			    (- 
			     (current-seconds) 
			     start-seconds)))))
	 (kill-tries 0))
    ;; (tests:set-full-meta-info #f test-id run-id (calc-minutes) work-area)
    ;; (tests:set-full-meta-info test-id run-id (calc-minutes) work-area)
    (tests:set-full-meta-info #f test-id run-id (calc-minutes) work-area 10 update-db: #t)

    (let loop ((minutes   (calc-minutes))
	       (cpu-load  (alist-ref 'adj-core-load (common:get-normalized-cpu-load #f)))
	       (disk-free (get-df (current-directory)))
               (last-sync (current-seconds)))
      ;; (common:telemetry-log "zombie" (conc "launch:monitor-job - top of loop encountered at "(current-seconds)" with last-sync="last-sync))
      ;; top of loop encountered at "(current-seconds)" with
      ;; last-sync="last-sync))
      (let* ((over-time     (> (current-seconds) (+ last-sync update-period)))
             (new-cpu-load  (let* ((load  (alist-ref 'adj-core-load (common:get-normalized-cpu-load #f)))
                                   (delta (abs (- load cpu-load))))
                              (if (> delta 0.1) ;; don't bother updating with small changes
                                  load
                                  #f)))
             (new-disk-free (let* ((df    (if over-time ;; only get df every 30 seconds

         ((test-get-kill-request run-id test-id)
          (set! kill-reason "KILLING TEST since received kill request (KILLREQ)")
          (set! kill-job? #t))
         ((and runtlim (> (- (current-seconds) start-seconds) runtlim))
          (set! kill-reason (conc "KILLING TEST DUE TO TIME LIMIT EXCEEDED! Runtime=" (- (current-seconds) start-seconds) " seconds, limit=" runtlim))
          (set! kill-job? #t))
         ((equal? status "DEAD")
          (tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f update-db: #t)
          (rmt:set-state-status-and-roll-up-items run-id test-id 'foo "RUNNING" "n/a" "was marked dead; really still running.")
          ;;(set! kill-reason "KILLING TEST because it was marked as DEAD by launch:handle-zombie-tests (might indicate really overloaded server or else overzealous setup.deadtime)") ;; MARK RUNNING
          (set! kill-job? #f)))

        (debug:print 4 *default-log-port* "cpu: " new-cpu-load " disk: " new-disk-free " last-sync: " last-sync " do-sync: " do-sync)
        (launch:handle-zombie-tests run-id)
        (if do-sync ;; save meta data about the running of this test



	    (tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f))

	  )

	(if kill-job? 
	    (begin
              (debug:print-info 0 *default-log-port* "proceeding to kill test: "kill-reason)
	      (mutex-lock! m)
	      ;; NOTE: The pid can change as different steps are run. Do we need handshaking between this
	      ;;       section and the runit section? Or add a loop that tries three times with a 1/4 second
	      ;;       between tries?

	    (begin
	      (thread-sleep! 3) ;; (+ 3 (random 6))) ;; add some jitter to the call home time to spread out the db accesses
	      (if (hash-table-ref/default misc-flags 'keep-going #f)  ;; keep originals for cpu-load and disk-free unless they change more than the allowed delta
		  (loop (calc-minutes)
                        (or new-cpu-load cpu-load)
                        (or new-disk-free disk-free)
                        (if do-sync (current-seconds) last-sync)))))))
    (tests:update-central-meta-info run-id test-id 
       (get-cpu-load)
       (get-df (current-directory))
       (calc-minutes) #f #f update-db: #t)) ;; NOTE: Checking twice for keep-going is intentional


(define (launch:execute encoded-cmd)
  (let* ((cmdinfo    (common:read-encoded-string encoded-cmd))
	 (tconfigreg #f))
    (setenv "MT_CMDINFO" encoded-cmd)
    ;;(bb-check-path msg: "launch:execute incoming")

                                                    (debug:print 0 *default-log-port* "INFO: missing files from " work-area ": " (string-intersperse bad-files ", "))
                                                    (launch:test-copy testpath work-area))))
                                            ;; one more time, change to the work-area directory
                                            (change-directory work-area)))
	       ) ;; let*

	  (if contour (setenv "MT_CONTOUR" contour))

	  ;; immediated set some key variables from CMDINFO data, yes, these will be set again below ...
	  ;;
	  (setenv "MT_TESTSUITENAME" areaname)
	  (setenv "MT_RUN_AREA_HOME" top-path)
	  (set! *toppath* top-path)
          (change-directory *toppath*) ;; temporarily switch to the run area home
	  (setenv "MT_TEST_RUN_DIR"  work-area)

				(db:test-get-host        test-info)
				(begin
				  (debug:print 0 *default-log-port* "ERROR: failed to find a record for test-id " test-id ", exiting.")
				  (exit))))
		 (test-pid  (db:test-get-process_id  test-info)))
	    (cond
             ;; -mrw- I'm removing KILLREQ from this list so that a test in KILLREQ state is treated as a "do not run" flag.
	     ;;((or (member (db:test-get-state test-info) '("INCOMPLETE" "KILLED" "UNKNOWN" "STUCK")) ;; prior run of this test didn't complete, go ahead and try to rerun
	     ;;	  (and (equal? (db:test-get-state test-info) "COMPLETED")                           ;; completed/abort => rerun if asked
	     ;;	       (member (db:test-get-status test-info) '("ABORT"))))
	     ((member (db:test-get-state test-info) '("INCOMPLETE" "KILLED" "UNKNOWN" "STUCK")) ;; prior run of this test didn't complete, go ahead and try to rerun
	      (debug:print 0 *default-log-port* "INFO: test is INCOMPLETE or KILLED, treat this execute call as a rerun request")
	      ;; (tests:test-force-state-status! run-id test-id "REMOTEHOSTSTART" "n/a")

              (rmt:general-call 'set-test-start-time #f test-id)
              (rmt:test-set-state-status run-id test-id "REMOTEHOSTSTART" "n/a" #f)
	      ) ;; prime it for running

			       	(debug:print 0 *default-log-port* "test " test-name "/" item-path " not completed")
              (if (not (null? tal))
		  (loop (car tal) (cdr tal)))))))))))

;; replaced below with version that does not ssh if checking on localhost
#;(define (launch:is-test-alive host pid)
  (if (and host pid (not (equal? host "n/a")))
      (let* ((is-local (equal? host (get-host-name)))
	     (ssh-cmd   (if is-local " " (conc "ssh " host " ")))
	     (cmd (conc ssh-cmd "pstree -A " pid))
	     (output (with-input-from-pipe cmd read-lines)))
	(debug:print 2 *default-log-port* "Running " cmd " received " output)
	(if (eq? (length output) 0)
	   #f
	   #t))
      #t))

           (else #f))))
    (when do-scan?
      (debug:print 1 *default-log-port* "INFO: search and mark zombie tests")
      (rmt:set-var key (current-seconds))
      (rmt:find-and-mark-incomplete run-id #f))))


(defstruct launch:ajt
  (vars '())
  (exekey #f)
  (host-type #f)
  (test-sig  #f)
  (cmdline   #f))

;; append vars
(define (launch:ajt-add-vars dat vars)
  (launch:ajt-vars-set! dat (append (launch:ajt-vars dat) vars)))

;; 1. look though disks list for disk with most space
;; 2. create run dir on disk, path name is meaningful
;; 3. create link from run dir to megatest runs area 
;; 4. remotely run the test on allocated host
;;    - could be ssh to host from hosts table (update regularly with load)
;;    - could be netbatch
;;      (launch-test db (cadr status) test-conf))
(define (launch-test test-id run-id run-info keyvals runname test-conf test-name test-path itemdat params)
  (mutex-lock! *launch-setup-mutex*) ;; setting variables and processing the testconfig is NOT thread-safe, reuse the launch-setup mutex
  (let* (;; locking code removed from here commented out and pasted at end of file












	 (item-path       (item-list->path itemdat))
	 (contour         #f)                         ;; NOT READY FOR THIS (args:get-arg "-contour")))
	 ;; launcher-mode will be 'adjutant or 'normal
	 (launcher-mode   (string->symbol (or (configf:lookup *configdat* "jobtools" "mode") "normal")))
	 (ajtdat          (make-launch:ajt)))
    (let loop ((delta        (- (current-seconds) *last-launch*))
	       (launch-delay (configf:lookup-number *configdat* "setup" "launch-delay" default: 0)))
      (if (> launch-delay delta)
	  (begin
	    (if (common:low-noise-print 1200 "test launch delay") ;; every two hours or so remind the user about launch delay.
		(debug:print-info 0 *default-log-port* "NOTE: test launches are delayed by " launch-delay " seconds. See megatest.config launch-delay setting to adjust.")) ;; launch of " test-name " for " (- launch-delay delta) " seconds"))
	    (thread-sleep! (- launch-delay delta))
	    (loop (- (current-seconds) *last-launch*) launch-delay))))
    (change-directory *toppath*)

    (let ((var-list (append
		     (list
		      (list "MT_RUN_AREA_HOME" *toppath*)
		      (list "MT_TEST_NAME" test-name)
		      (list "MT_RUNNAME"   runname)
		      (list "MT_ITEMPATH"  item-path)
		      (list "MT_CONTOUR"   contour)
		      )
		     itemdat)))
       ;; consolidate this code with the code in megatest.scm for
       ;; "-execute", *maybe* - the longer they are set the longer
       ;; each launch takes (must be non-overlapping with the vars)
      (alist->env-vars var-list)
      ;; the var-list into the ajtdat adjutant record whether it is needed or not.
      (launch:ajt-add-vars ajtdat var-list))
    
    (let* ((tregistry       (tests:get-all)) ;; third param (below) is system-allowed
           ;; for tconfig, why do we allow fallback to test-conf?
	   (tconfig         (or (tests:get-testconfig test-name item-path tregistry #t force-create: #t)
				(begin
                                  (debug:print 0 *default-log-port* "WARNING: falling back to pre-calculated testconfig. This is likely not desired.")
                                  test-conf))) ;; force re-read now that all vars are set
	   (useshell        (let ((ush (configf:lookup *configdat* "jobtools"     "useshell")))

	   (subrun          (> (length (hash-table-ref/default tconfig "subrun"  '())) 0)) ;; send a flag to process a subrun
	   ;; (diskspace       (configf:lookup tconfig   "requirements" "diskspace"))
	   ;; (memory          (configf:lookup tconfig   "requirements" "memory"))
	   ;; (hosts           (configf:lookup *configdat* "jobtools"     "workhosts")) ;; I'm pretty sure this was never completed
	   (remote-megatest (configf:lookup *configdat* "setup" "executable"))
	   (run-time-limit  (or (configf:lookup  tconfig   "requirements" "runtimelim")
				(configf:lookup  *configdat* "setup" "runtimelim")))




	   (local-megatest  (common:find-local-megatest))



	   (launcher        (let ((l (common:get-launcher *configdat* test-name item-path launcher-mode)))
			      (if (string? l)
				  (string-split l)





				  l))) ;; some nonhomogenuity here. '(cmd param1 param2 ...) OR '(host-type launcher)
	    ;; (item-list->path itemdat))) ;; test-path is the full path including the item-path
	   (test-sig        (conc (common:get-testsuite-name) ":" test-name ":" item-path))
	   (work-area       #f)
	   (toptest-work-area #f) ;; for iterated tests the top test contains data relevant for all
	   (diskpath   #f)
	   (cmdparms   #f)
	   (fullcmd    #f) ;; (define a (with-output-to-string (lambda ()(write x))))
	   (mt-bindir-path #f)
	   (testinfo   (rmt:get-test-info-by-id run-id test-id))
	   (mt_target  (string-intersperse (map cadr keyvals) "/"))
	   (debug-param (append (if (args:get-arg "-debug")  (list "-debug" (args:get-arg "-debug")) '())
				(if (args:get-arg "-logging")(list "-logging") '())
				(if (configf:lookup *configdat* "misc" "profilesw")
				    (list (configf:lookup *configdat* "misc" "profilesw"))
				    '()))))
      ;; save the test-sig in the ajtdat record
      (launch:ajt-test-sig-set! ajtdat test-sig)
      ;; go ahead and figure out if we have a host-type from the
      ;; launcher call above and save it in the ajtdat record
      (if (and (eq? launcher-mode 'adjutant)
	       (list? launcher)
	       (> (length launcher) 1))
	  (launch:ajt-host-type-set! ajtdat (car launcher)))
 
      ;; (if hosts (set! hosts (string-split hosts)))
      ;; set the megatest to be called on the remote host
      (if (not remote-megatest)(set! remote-megatest local-megatest)) ;; "megatest"))
      (set! mt-bindir-path (pathname-directory remote-megatest))
      ;; (if launcher (set! launcher (string-split launcher)))           ;; yuk!
      ;; set up the run work area for this test
      (if (and (args:get-arg "-preclean") ;; user has requested to preclean for this run
	       (not (member (db:test-get-rundir testinfo)(list "n/a" "/tmp/badname")))) ;; n/a is a placeholder and thus not a read dir
	  (begin
	    (debug:print-info 0 *default-log-port* "attempting to preclean directory " (db:test-get-rundir testinfo) " for test " test-name "/" item-path)
	    (runs:remove-test-directory testinfo 'remove-data-only))) ;; remove data only, do not perturb the record
      

					(list 'runtlim   (if run-time-limit (common:hms-string->seconds run-time-limit) #f))
					(list 'env-ovrd  (hash-table-ref/default *configdat* "env-override" '())) 
					(list 'set-vars  (if params (hash-table-ref/default params "-setvars" #f)))
					(list 'runname   runname)
					(list 'mt-bindir-path mt-bindir-path))))))))

        (setenv "MT_CMDINFO" cmdparms)  ;; setting this for use in nblauncher
        
      ;; save the cmdparms in the ajtdat
      (launch:ajt-exekey-set! ajtdat cmdparms)
      
      ;; clean out step records from previous run if they exist
      ;; (rmt:delete-test-step-records run-id test-id)
      ;; if the dir does not exist we may have a itempath where individual variables are a path, launch anyway
      (if (common:file-exists? work-area)
	  (change-directory work-area)) ;; so that log files from the launch process don't clutter the test dir

      ;; save the command line for adjutant mode (might never be needed but best to assemble it here)
      (launch:ajt-cmdline-set! ajtdat (string-intersperse
				       (append (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param)))

      (cond       
       (launcher
	(set! fullcmd (append launcher (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param)))

       (else
	(if (not useshell)(debug:print 0 *default-log-port* "WARNING: internal launching will not work well without \"useshell yes\" in your [jobtools] section"))
	(set! fullcmd (append (list remote-megatest "-m" test-sig "-execute" cmdparms) debug-param (list (if useshell "&" ""))))))
      
      (if (args:get-arg "-xterm")(set! fullcmd (append fullcmd (list "-xterm"))))
      (debug:print 1 *default-log-port* "Launching " work-area)
      ;; set pre-launch-env-vars before launching, keep the vars in prevvals and put the envionment back when done
      (debug:print 4 *default-log-port* "fullcmd: " fullcmd)
      (set! *last-launch* (current-seconds)) ;; all that junk above takes time, set this as late as possible.

      (let* ((env-override-vars  (hash-table-ref/default *configdat* "env-override" '()))
	     (commonprevvals     (alist->env-vars env-override-vars))
	     (misc-vars          (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)
					       (list "MT_ITEMPATH"  item-path))

					 itemdat))
	     (miscprevvals   (alist->env-vars misc-vars));; consolidate this code with the code in megatest.scm for "-execute"
	     (test-vars      (hash-table-ref/default tconfig "pre-launch-env-overrides" '()))
	     (testprevvals   (alist->env-vars test-vars))
			      
	     ;; Launchwait defaults to true, must override it to turn off wait
	     (launchwait     (if (equal? (configf:lookup *configdat* "setup" "launchwait") "no") #f #t))
	     ;; BB: TODO: refactor this to examine return code of launcher, if nonzero, set state to launch failed.
	     (launch-results-prev (if (eq? launcher-mode 'adjutant)
				      '(#t 0) ;; just some fake data to fool downstream but non-applicable code
				      (apply (if launchwait
						 process:cmd-run-with-stderr-and-exitcode->list
						 process-run)
					     (if useshell
						 (let ((cmdstr (string-intersperse fullcmd " ")))
						   (if launchwait
						       cmdstr
						       (conc cmdstr " >> mt_launch.log 2>&1 &")))
						 (car fullcmd))
					     (if useshell
						 '()
						 (cdr fullcmd)))))
             (success        (if launchwait (equal? 0 (cadr launch-results-prev)) #t))
             (launch-results (if launchwait (car launch-results-prev) launch-results-prev)))

	(launch:ajt-add-vars ajtdat env-override-vars)
	(launch:ajt-add-vars ajtdat misc-vars)
	(launch:ajt-add-vars ajtdat test-vars)

	;; if in adjutant mode we register the job in the jobs_queue
	;; then fire off an adjutant runner
	;;
	(if (eq? launcher-mode 'adjutant)
	    (let* ((adjutant-runner-cmd (append (cdr launcher)
						(list remote-megatest "-adjutant"
						      (launch:ajt-host-type ajtdat)
						      "-start-dir" *toppath*)))
		   (adj-cmd     (conc (string-intersperse (map conc adjutant-runner-cmd) " ")
				      "&")))         
	      (rmt:no-sync-add-job
	       (launch:ajt-host-type  ajtdat)
	       (launch:ajt-vars ajtdat)
	       (launch:ajt-exekey     ajtdat)
	       (launch:ajt-cmdline    ajtdat))
	      (print "adj-cmd: " adj-cmd)
	      (system adj-cmd)
	      ))
	
	(if (not success)
            (tests:test-set-status! run-id test-id "COMPLETED" "DEAD" "launcher failed; exited non-zero; check mt_launch.log" #f)) ;; (if launch-results launch-results "FAILED"))

	;; (rmt:no-sync-del! lock-key)         ;; release the lock for starting this test
	(if (not launchwait) ;; give the OS a little time to allow the process to start
	    (thread-sleep! 0.01))
	(with-output-to-file "mt_launch.log"
	  (lambda ()
	    (print "LAUNCHCMD: " (string-intersperse fullcmd " "))
	    (if (list? launch-results)

	      ;; but this hack will work! Thanks go to Alan Post of the Chicken email list
	      ;; NB// Is this still needed? Should be safe to go back to "exit" now?
	      (process-signal (current-process-id) signal/kill)
	      ))
	(alist->env-vars miscprevvals)
	(alist->env-vars testprevvals)
	(alist->env-vars commonprevvals)
	;; yes, really should mutex all the way to here. Need to put this entire process into a fork.
	;; the unlock previously was further up. This seemed wrong as we should not proceed until the
	;; vars have been reset.
	(mutex-unlock! *launch-setup-mutex*)
	launch-results))
    (change-directory *toppath*)
    (thread-sleep! (configf:lookup-number *configdat* "setup" "inter-test-delay" default: 0.0))))

;; recover a test where the top controlling mtest may have died
;;
(define (launch:recover-test run-id test-id)

		     (read-symbolic-link (conc "/proc/" pid "/cwd"))
		     #f)))
    ;; now wait on that process if all is correct
    ;; periodically update the db with runtime
    ;; when the process exits look at the db, if still RUNNING after 10 seconds set
    ;; state/status appropriately
    (process-wait pid)))


 ;; (lock-key        (conc "test-" test-id))
	;; (got-lock        (let loop ((lock        (rmt:no-sync-get-lock lock-key))
	;; 			     (expire-time (+ (current-seconds) 15))) ;; give up on getting the lock and steal it after 15 seconds
	;; 		    (if (car lock)
	;; 			#t
	;; 			(if (> (current-seconds) expire-time)
	;; 			    (begin
	;; 			      (debug:print-info 0 *default-log-port* "Timed out waiting for a lock to launch test " keyvals " " runname " " test-name " " test-path)
	;; 			      (rmt:no-sync-del! lock-key) ;; destroy the lock
	;; 			      (loop (rmt:no-sync-get-lock lock-key) expire-time)) ;; 
	;; 			    (begin
	;; 			      (thread-sleep! 1)
	;; 			      (loop (rmt:no-sync-get-lock lock-key) expire-time))))))