Megatest

           process.scm runs.scm tasks.scm tests.scm genexample.scm	\
           http-transport.scm filedb.scm tdb.scm client.scm mt.scm	\
           ezsteps.scm lock-queue.scm sdb.scm rmt.scm api.scm		\
           subrun.scm portlogger.scm archive.scm env.scm		\
           diff-report.scm cgisetup/models/pgdb.scm

# module source files
MSRCFILES = adjutant.scm mutils.scm mttop.scm
# ftail.scm rmtmod.scm commonmod.scm removed
# MSRCFILES = ducttape-lib.scm pkts.scm stml2.scm cookie.scm mutils.scm	\
#             mtargs.scm commonmod.scm dbmod.scm adjutant.scm ulex.scm	\
#             rmtmod.scm apimod.scm

GUISRCF = dashboard-context-menu.scm dashboard-tests.scm		\
          dashboard-guimonitor.scm gutils.scm dcommon.scm tree.scm	\

(declare (unit adjutant))

(module adjutant *

(import scheme chicken data-structures extras files)
(import (prefix sqlite3 sqlite3:) posix typed-records srfi-18 srfi-69
	md5 message-digest matchable
	regex srfi-1)

(define (adjutant-run host-type rmt:no-sync-take-job)
  (print "Running the adjutant!")
  (let loop ((wait-count 0))
    (if (< wait-count 10) ;; 6 x 10 seconds = one minute
	(let* ((dat (rmt:no-sync-take-job host-type)))
	  (match dat
	    ((id ht vars exekey cmdline state event-time last-update)
	     (system cmdline)
	     (loop 0))
	    (else
	     (thread-sleep! 10)
	     (loop (+ wait-count 1)))))
	(print "I'm bored. Exiting."))))

)

    ((not (vector? dat))                    ;; it is an error to not receive a vector
     (vector #f (vector #f "remote must be called with a vector")))
    ((> *api-process-request-count* 20) ;; 20)
     (debug:print 0 *default-log-port* "WARNING: api:execute-requests received an overloaded message.")
     (set! *server-overloaded* #t)
     (vector #f (vector #f 'overloaded))) ;; the inner vector is what gets returned. nope, don't know why. please refactor!
    (else  
     (let* ((cmd-in            (common:safe-vector-ref dat 0 'nocmd))
            (cmd               (if (symbol? cmd-in)
				   cmd-in
				   (string->symbol cmd-in)))
            (params            (common:safe-vector-ref dat 1 '()))
            (start-t           (current-milliseconds))
            (readonly-mode     (dbr:dbstruct-read-only dbstruct))
            (readonly-command  (member cmd api:read-only-queries))
            (writecmd-in-readonly-mode (and readonly-mode (not readonly-command)))
            (foo               (begin
                                 #;(common:telemetry-log (conc "api-in:"(->string cmd))
                                                       payload: `((params . ,params)))
                                 
                                 #t))
            (res    
             (if writecmd-in-readonly-mode
                 (conc "attempt to run write command "cmd" on a read-only database")
                 (case cmd
                   ;;===============================================
                   ;; READ/WRITE QUERIES
                   ;;===============================================

		   ((nocmd)                         '(#f "All broken!"))

                   ((get-keys-write)                        (db:get-keys dbstruct)) ;; force a dummy "write" query to force server; for debug in -repl
                   
                   ;; SERVERS
                   ((start-server)                    (apply server:kind-run params))
                   ((kill-server)                     (set! *server-run* #f))

                   ((tasks-get-last)            (apply tasks:get-last dbstruct params))

		   ;; NO SYNC DB
		   ((no-sync-set)               (apply db:no-sync-set         *no-sync-db* params))
		   ((no-sync-get/default)       (apply db:no-sync-get/default *no-sync-db* params))
		   ((no-sync-del!)              (apply db:no-sync-del!        *no-sync-db* params))
		   ((no-sync-get-lock)          (apply db:no-sync-get-lock    *no-sync-db* params))
		   ((no-sync-add-job)           (apply db:no-sync-add-job     *no-sync-db* params))
		   ((no-sync-take-job)          (apply db:no-sync-take-job    *no-sync-db* params))
		   ((no-sync-job-records-clean) (apply db:no-sync-job-records-clean *no-sync-db* params))
		   
                   ;; ARCHIVES
                   ;; ((archive-get-allocations)   
                   ((archive-register-disk)     (apply db:archive-register-disk dbstruct params))
                   ((archive-register-block-name)(apply db:archive-register-block-name dbstruct params))
                   ;; ((archive-allocate-testsuite/area-to-block)(apply db:archive-allocate-testsuite/area-to-block dbstruct block-id testsuite-name areakey))
		   
                   ;;======================================================================
                   ;; READ ONLY QUERIES
                   ;;======================================================================

                   ;; KEYS
                   ((get-key-val-pairs)               (apply db:get-key-val-pairs dbstruct params))
                   ((get-keys)                        (db:get-keys dbstruct))

       ;; save all stats
       (let ((delta-t (- (current-milliseconds)
			 start-t)))
	 (hash-table-set! *db-api-call-time* cmd
			  (cons delta-t (hash-table-ref/default *db-api-call-time* cmd '()))))
       (if writecmd-in-readonly-mode
           (begin
             #;(common:telemetry-log (conc "api-out:"(->string cmd))
                                   payload: `((params . ,params)
                                              (ok-res . #t)))
	     (vector #f res))
           (begin
             #;(common:telemetry-log (conc "api-out:"(->string cmd))
                                   payload: `((params . ,params)
                                              (ok-res . #f)))
             (vector #t res))))))))

;; http-server  send-response
;;                 api:process-request
;;                    db:*
;;
;; NB// Runs on the server as part of the server loop
;;
(define (api:process-request dbstruct $) ;; the $ is the request vars proc
  (set! *api-process-request-count* (+ *api-process-request-count* 1))
  (let* ((cmd     ($ 'cmd))
	 (paramsj ($ 'params))
	 (params  (db:string->obj paramsj transport: 'http)) ;; incoming data from the POST (or is it a GET?)
	 (resdat  (api:execute-requests dbstruct (vector cmd params))) ;; process the request, resdat = #( flag result )
	 (success (common:safe-vector-ref resdat 0 #f))
	 (res     (common:safe-vector-ref resdat 1 #f))) ;; (vector flag payload), get the payload, ignore the flag (why?)
    (if (not success)
	(debug:print 0 *default-log-port* "ERROR: success flag is #f for " cmd " with params " params))
    (if (> *api-process-request-count* *max-api-process-requests*)
	(set! *max-api-process-requests* *api-process-request-count*))
    (set! *api-process-request-count* (- *api-process-request-count* 1))
    ;; This can be here but needs controls to ensure it doesn't run more than every 4 seconds
    ;; (rmt:dat->json-str

        ;;     copy <file>.hrs.gz <file>.days.gz
        (when (>= (age-wks daysfile) 1)
          (copy daysfile wksfile)
          (copy hrsfile daysfile))
        #t)
      #f))
  
(define (common:safe-vector-ref vec indx default)
  (if (vector? vec)
      (handle-exceptions
	  exn
	(begin
	  (debug:print-info 0 *default-log-port* "remote data issue: exn=" exn)
	  default)
	(vector-ref vec indx))
      default))

        
;; Rotate logs, logic: 
;;                 if > 500k and older than 1 week:
;;                     remove previous compressed log and compress this log
;; WARNING: This proc operates assuming that it is in the directory above the
;;          logs directory you wish to log-rotate.
;;

;; 
;; [hosts]
;; arm cubie01 cubie02
;; x86_64 zeus xena myth01
;; allhosts #{g hosts arm} #{g hosts x86_64}
;; 
;; [host-types]
;; C/M/A      lets megatest know this launcher provides C cores, M bytes memory for architecture A 
;; 2/2G/arm   smart -cores 2 -memory 2G -arch arm
;; general #MTLOWESTLOAD #{g hosts allhosts}
;; arm     #MTLOWESTLOAD #{g hosts arm}
;; nbgeneral nbjob run JOBCOMMAND -log $MT_LINKTREE/$MT_TARGET/$MT_RUNNAME.$MT_TESTNAME-$MT_ITEM_PATH.lgo
;;
;; NOTE: host-rules is ONLY used for MTLOWESTLOAD
;;
;; [host-rules]
;; # maxnload   => max normalized load
;; # maxnjobs   => max jobs per cpu
;; # maxjobrate => max jobs per second
;; general maxnload=1.1; maxnjobs=1.2; maxjobrate=0.1 
;; 
;; [launchers]
;; envsetup general
;; xor/%/n 2/2G/arm
;; % nbgeneral
;; 
;; [jobtools]
;; # if defined and not "no" flexi-launcher will bypass "launcher" unless no match.
;; flexi-launcher yes  
;; launcher nbfake
;; mode adjutant|normal (default is normal)
;; 
;;
;; mode is 'normal (i.e. directly use launcher) or 'adjutant (i.e. use adjutant)
;;
(define (common:get-launcher configdat testname itempath mode)
  (let ((fallback-launcher (configf:lookup configdat "jobtools" "launcher")))
    (if (and (configf:lookup configdat "jobtools" "flexi-launcher") ;; overrides launcher
	     (not (equal? (configf:lookup configdat "jobtools" "flexi-launcher") "no")))
	(let* ((launchers         (hash-table-ref/default configdat "launchers" '())))
	  (if (null? launchers)
	      fallback-launcher
	      (let loop ((hed (car launchers))
			 (tal (cdr launchers)))
		(let ((patt      (car hed))
		      (host-type (cadr hed)))
		  (if (tests:match patt testname itempath) ;; have a launcher match for this test
		      (begin
			(debug:print-info 2 *default-log-port* "Have flexi-launcher match for " testname "/" itempath " = " host-type)
			(let ((launcher (configf:lookup configdat "host-types" host-type))) ;; find the actual launcher from the host-types table
			  ;; if we are in adjutant mode then we want to return both host-type and launcher
			  (if launcher
			      (let* ((launcher-parts (string-split launcher))
				     (launcher-exe   (car launcher-parts)))
				(if (equal? launcher-exe "#MTLOWESTLOAD") ;; this is our special case, we will find the lowest load and craft a nbfake commandline
				    (let host-loop ((targ-host (common:get-least-loaded-host (cdr launcher-parts) host-type configdat))
						    (count     100))
				      (if targ-host
					  (conc "remrun " targ-host)
					  (if (> count 0)
					      
					      (begin
						(debug:print 0 *default-log-port* "INFO: Waiting for a host for host-type " host-type)
						(thread-sleep! (- 101 count))
						(host-loop (common:get-least-loaded-host (cdr launcher-parts) host-type configdat)
							   (- count 1)))
					      (begin
						(debug:print 0 *default-log-port* "FATAL: Failed to find a host from #MTLOWESTLOAD for host-type " host-type)
						(exit)))))
				    (case mode
				      ((adjutant) (list host-type launcher))
				      (else       launcher))))
			      (begin
				(debug:print-info 0 *default-log-port* "WARNING: no launcher found for host-type " host-type)
				(if (null? tal)
				    fallback-launcher
				    (loop (car tal)(cdr tal)))))))
		      ;; no match, try again
		      (if (null? tal)

    ht))

;; if 
(define (configf:read-alist fname)
  (handle-exceptions
      exn
    (begin
      (debug:print-info 0 *default-log-port* "unable to read alist " fname ". exn=" exn)
      #f)
    (configf:alist->config
     (with-input-from-file fname read))))

(define (configf:write-alist cdat fname)
  (if (not (common:faux-lock fname))
      (debug:print 0 *default-log-port* "INFO: Could not get lock on " fname))

  (let* ((dbpath (db:dbfile-path))
	 (dbname (conc dbpath "/no-sync.db"))
	 (db-exists (common:file-exists? dbname))
	 (db     (sqlite3:open-database dbname)))
    (sqlite3:set-busy-handler! db (sqlite3:make-busy-timeout 136000))
    (if (not db-exists)
	(begin
	  (sqlite3:execute db "CREATE TABLE IF NOT EXISTS no_sync_metadat (var TEXT,val TEXT, CONSTRAINT no_sync_metadat_constraint UNIQUE (var));")
	  (sqlite3:execute db "PRAGMA journal_mode=WAL;")))
    ;; MOVE THIS TABLE CREATION TO THE (begin above in about six months (it is Sep 2020 right now).
    (sqlite3:execute db "CREATE TABLE IF NOT EXISTS jobs_queue (id INTEGER PRIMARY KEY, host_type TEXT, cores INTEGER, memory TEXT, vars TEXT, exekey TEXT, cmdline TEXT, state TEXT, event_time INTEGER, last_update INTEGER);")
    ;; not sure I'll use this next one. I prefer if tests simply append to a file:
    ;;    last-update-seconds cpuload tmpspace rundirspace
    (sqlite3:execute db "CREATE TABLE IF NOT EXISTS test_extra_data (id INTEGER PRIMARY KEY, run_id INTEGER, test_id INTEGER, last_seen_running INTEGER);") 
    (sqlite3:execute db "PRAGMA synchronous = 0;")
    db))

(define (db:no-sync-add-job db-in host-type vars-list exekey cmdline)
  (sqlite3:execute (db:no-sync-db db-in) "INSERT INTO jobs_queue (host_type,vars,exekey,cmdline,state,event_time,last_update) VALUES (?,?,?,?,?,?,?);"
		   host-type
		   (with-output-to-string
		     (lambda ()
		       (write vars-list)))
		   exekey cmdline "waiting" (current-seconds)(current-seconds)))

;; find next job (waiting longest) that matches host-type - future, we'll find jobs that fit if no exact match
(define (db:no-sync-take-job db-in host-type)
  (let* ((db   (db:no-sync-db db-in))
	 (stmt1 "SELECT id,host_type,vars,exekey,cmdline,state,event_time,last_update FROM jobs_queue WHERE host_type=? AND state != 'taken' ORDER BY event_time ASC;")
	 (stmt1h (sqlite3:prepare db stmt1))
	 (stmt2  "UPDATE jobs_queue SET state='taken',last_update=? WHERE id=?;")
	 (stmt2h (sqlite3:prepare db stmt2))
	 (res    (sqlite3:with-transaction
		  db
		  (lambda ()
		    (let* ((matching-jobs (sqlite3:fold-row
					   (lambda (res . row) ;; id host-type vars exekey state event-time last-update)
					     (cons row res))
					   '()
					   stmt1h
					   host-type)))
		      (if (null? matching-jobs)
			  #f
			  (let ((choosen-one  (let loop ((tal matching-jobs)
							 (res #f)) ;; put bestest one in here
						(if (null? tal)
						    res
						    (let ((curr (car tal))
							  (rem  (cdr tal)))
						      curr) ;; here we will compare with res, if better candidate the loop with curr else loop with res
						    ))))
			    (if choosen-one ;; we need to mark it as taken
				(sqlite3:execute stmt2h (current-seconds) (car choosen-one)))
			    choosen-one)))))))
    (sqlite3:finalize! stmt1h) ;; it'd be nice to cache these and finalize on exit.
    (sqlite3:finalize! stmt2h)
    res))

;; clean out old jobs in queue, i.e. taken and event_time > 24 hrs ago
;;
(define (db:no-sync-job-records-clean db)
  (sqlite3:execute (db:no-sync-db db) "DELETE FROM jobs_queue WHERE state='taken' AND event_time < ?;" (- (current-seconds)(* 24 3600))))
		 

;; if we are not a server create a db handle. this is not finalized
;; so watch for problems. I'm still not clear if it is needed to manually
;; finalize sqlite3 dbs with the sqlite3 egg.
;;
(define (db:no-sync-db db-in)
  (mutex-lock! *db-access-mutex*)

(define (db:no-sync-set db var val)
  (sqlite3:execute (db:no-sync-db db) "INSERT OR REPLACE INTO no_sync_metadat (var,val) VALUES (?,?);" var val))

(define (db:no-sync-del! db var)
  (sqlite3:execute (db:no-sync-db db) "DELETE FROM no_sync_metadat WHERE var=?;" var))

(define (db:no-sync-get/default db-in var default)
  (let ((db  (db:no-sync-db db-in))
	(res default))
    (sqlite3:for-each-row
     (lambda (val)
       (set! res val))
     (db:no-sync-db db)
     "SELECT val FROM no_sync_metadat WHERE var=?;"
     var)
    (if res

                 (item-path (vector-ref running-test 11)))
			       	(debug:print 0 *default-log-port* "test " test-name "/" item-path " not completed")
              (if (not (null? tal))
				  (loop (car tal) (cdr tal)))))))))))        
 
(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 'target    mt_target)
					(list 'contour   contour)
					(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))))))))
      ;; 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))))))
	 

(declare (uses tdb))
(declare (uses mt))
(declare (uses api))
(declare (uses tasks)) ;; only used for debugging.
(declare (uses env))
(declare (uses diff-report))

(declare (uses adjutant))
(import adjutant)

(declare (uses mttop))
(import mttop)

;; (declare (uses ftail))
;; (import ftail)

(define *db* #f) ;; this is only for the repl, do not use in general!!!!

(include "common_records.scm")
(include "key_records.scm")
(include "db_records.scm")
(include "run_records.scm")
(include "megatest-fossil-hash.scm")

(use (prefix sqlite3 sqlite3:) srfi-1 posix regex regex-case srfi-69 (prefix base64 base64:)
     readline apropos json http-client directory-utils typed-records matchable
     http-client srfi-18 extras format call-with-environment-variables)

;; Added for csv stuff - will be removed
;;
(use sparse-vectors)

(require-library mutils)

  version " megatest-version "
  license GPL, Copyright Matt Welland 2006-2017
 
Usage: megatest [options]
  -h                      : this help
  -manual                 : show the Megatest user manual
  -version                : print megatest version (currently " megatest-version ")
  help                    : help for the new Megatest interface

Launching and managing runs
  -run                    : run all tests or as specified by -testpatt
  -remove-runs            : remove the data for a run, requires -runname and -testpatt
                            Optionally use :state and :status, use -keep-records to remove only
                            the run data. Use -kill-wait to override the 10 second
                            per test wait after kill delay (e.g. -kill-wait 0). 

  -sync-to-megatest.db    : pull data from cache files in /tmp/$USER to megatest.db
  -sync-to dest           : sync to new postgresql central style database
  -update-meta            : update the tests metadata for all tests
  -setvars VAR1=val1,VAR2=val2 : Add environment variables to a run NB// these are
                                 overwritten by values set in config files.
  -server -|hostname      : start the server (reduces contention on megatest.db), use
                            - to automatically figure out hostname
  -adjutant host-type     : start the server/adjutant with given host-type
                            use 0,0 to auto use full machine
  -transport http|rpc     : use http or rpc for transport (default is http) 
  -log logfile            : send stdout and stderr to logfile
  -list-servers           : list the servers 
  -kill-servers           : kill all servers
  -repl                   : start a repl (useful for extending megatest)
  -load file.scm          : load and run file.scm

Called as " (string-intersperse (argv) " ") "
Version " megatest-version ", built from " megatest-fossil-hash ))

;;  -gui                    : start a gui interface
;;  -config fname           : override the runconfigs file with fname


(mttop-run (command-line-arguments)
	   '("help"))

;; process args
(define remargs (args:get-args 
		 (argv)
		 (list  "-runtests"  ;; run a specific test
			"-config"    ;; override the config file name
			"-append-config"
			"-execute"   ;; run the command encoded in the base64 parameter

;;
(if (args:get-arg "-server")
    (let ((tl        (launch:setup))
          (transport-type (string->symbol (or (args:get-arg "-transport") "http"))))
      (server:launch 0 transport-type)
      (set! *didsomething* #t)))


(define (naylist->alist inlst)
  (map (lambda (dat)
	 (cons (car dat)
	       (or (if (list?   (cdr dat))
		       (if (null? (cdr dat)) ""
			   (cadr dat))
		       (cdr dat))
		   ""))) ;; we need a string for call-with-environment-variables
       inlst))

;; The adjutant is a bit different, it does NOT run (launch:setup) as it is not necessarily tied to
;; a specific Megatest area. Detail are being hashed out and this may change.
;;
(if (args:get-arg "-adjutant")
    (let* ((host-type (args:get-arg "-adjutant")))
      (launch:setup) ;; dang it, wish this wasn't needed
      (print "Running the adjutant!")
      (let loop ((wait-count 0))
	(if (< wait-count 10) ;; 6 x 10 seconds = one minute
	    (let* ((dat (rmt:no-sync-take-job host-type)))
	      (match dat
		  ((id ht vars exekey cmdline state event-time last-update)
		      (let ((vars-alist (with-input-from-string vars read)
					))
			(print "Vars:")
			(pp vars-alist)
			(call-with-environment-variables
			 (naylist->alist vars-alist)
			 (lambda ()
			   (system cmdline))))
		      (loop 0))
		  (else
		   (thread-sleep! 10)
		   (loop (+ wait-count 1)))))
	    (print "I'm bored. Exiting.")))
      ;; (adjutant-run (args:get-arg "-ajutant") rmt:no-sync-take-job)
      (set! *didsomething* #t)))

(if (or (args:get-arg "-list-servers")
        (args:get-arg "-kill-servers"))
    (let ((tl (launch:setup)))
      (if tl ;; all roads from here exit
	  (let* ((servers (server:get-list *toppath*))

;; Copyright 2006-2011, 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.

;; This is from the perl world, a hash of hashes is a super easy way to keep a handle on
;; lots of disparate data
;;

(declare (unit mttop))

(module mttop
    *

(import chicken scheme
	;; data-structures posix
	srfi-1
	;; srfi-13
	srfi-69
	ports
	extras
	regex
	posix
	data-structures
	matchable
	)

(define (str-is-cmd cmd all-cmds)
  (let* ((rx  (regexp (conc "^" cmd ".*")))
	 (mx  (filter string? (map (lambda (x)
				     (let ((res (string-match rx x)))
				       (if res (car res) #f)))
				   all-cmds))))
    (if (eq? (length mx) 1) ;; have a command
	(car mx)
	#f)))

(define (mttop-run args all-cmds)
  ;; any path through this call must end in exit if it is NOT an old Megatest call
  (if (null? args)
      #f ;; continue on and do the old Megatest stuff
      (let ((cmd (str-is-cmd (car args) all-cmds)))
	(if cmd
	    (begin
	      (case (string->symbol cmd)
		((help)(print "New help"))
		(else (print "Command " cmd " is not implemented yet.")))
	      (exit)) ;; always exit here
	    #f))))    ;; or continue on to Megatest old stuff here
  
)

	      cinfo
	      (if (server:check-if-running areapath)
		  (client:setup areapath)
		  #f))))

(define *send-receive-mutex* (make-mutex)) ;; should have separate mutex per run-id

(define *rmt-query-last-call-time* 0)
(define *rmt-query-last-rest-time* 0) ;; last time there was at least a 1/2 second rest - giving other processes access to the db

;; NOTE: This query rest algorythm will not adapt to long query times. REDESIGN NEEDED. TODO. FIXME.
;;
(define (rmt:query-rest)
  (let* ((now (current-milliseconds)))
    (cond
     ((> (- now *rmt-query-last-call-time*) 500)  ;; it's been a while since last query - no need to rest
      (set! *rmt-query-last-rest-time*  now)
      (set! *rmt-query-last-call-time*  now))
     ((> (- now *rmt-query-last-rest-time*) 5000) ;; no natural rests have happened
      (debug:print 0 *default-log-port* "query rest needed. blocking for 1/2 second.")
      (thread-sleep! 0.5) ;; force a rest of a half second
      (set! *rmt-query-last-rest-time* now)
      (set! *rmt-query-last-call-time* now))
     (else ;; sufficient rests have occurred, just record the last query time
      (set! *rmt-query-last-call-time* now)))))

;; RA => e.g. usage (rmt:send-receive 'get-var #f (list varname))
;;
(define (rmt:send-receive cmd rid params #!key (attemptnum 1)(area-dat #f)) ;; start attemptnum at 1 so the modulo below works as expected

  #;(common:telemetry-log (conc "rmt:"(->string cmd))
                        payload: `((rid . ,rid)
                                   (params . ,params)))
  (if (not (equal? (configf:lookup *configdat* "setup" "query-rest") "no"))
      (rmt:query-rest))
  
  (if (> attemptnum 2)
      (debug:print 0 *default-log-port* "INFO: attemptnum in rmt:send-receive is " attemptnum))
    
  (cond
   ((> attemptnum 2) (thread-sleep! 0.05))
   ((> attemptnum 10) (thread-sleep! 0.5))
   ((> attemptnum 20) (thread-sleep! 1)))

				  (vector #t '())) ;; should always get a vector but if something goes wrong return a dummy
				(if (and (vector? v)
					 (> (vector-length v) 1))
				    (let ((newvec (vector (vector-ref v 0)(vector-ref v 1))))
				      newvec)           ;; by copying the vector while inside the error handler we should force the detection of a corrupted record
				    (vector #t '()))))  ;; we could also check that the returned types are valid
			     (vector #t '())))
	 (success        (common:safe-vector-ref resdat 0 #f)) ;; (vector-ref resdat 0))
	 (res            (common:safe-vector-ref resdat 1 #f)) ;; (vector-ref resdat 1))
	 (duration       (- (current-milliseconds) start)))
    (if (and read-only qry-is-write)
        (debug:print 0 *default-log-port* "ERROR: attempt to write to read-only database ignored. cmd=" cmd))
    (if (not success)
	(if (> remretries 0)
	    (begin
	      (debug:print-error 0 *default-log-port* "local query failed. Trying again.")
	      (thread-sleep! (/ (random 5000) 1000)) ;; some random delay 
	      (rmt:open-qry-close-locally cmd run-id params remretries: (- remretries 1)))
	    (begin
	      (debug:print-error 0 *default-log-port* "too many retries in rmt:open-qry-close-locally, giving up")
	      #f))
	(begin
	  ;; (rmt:update-db-stats run-id cmd params duration)
	  ;; mark this run as dirty if this was a write, the watchdog is responsible for syncing it
	  (if qry-is-write
	      (let ((start-time (current-seconds)))
		(mutex-lock! *db-multi-sync-mutex*)
		(set! *db-last-access* start-time)  ;; THIS IS PROBABLY USELESS? (we are on a client)
                (mutex-unlock! *db-multi-sync-mutex*)))))
    res))

(define (rmt:send-receive-no-auto-client-setup connection-info cmd run-id params)
  (let* ((run-id   (if run-id run-id 0))
	 (res  	   (handle-exceptions
		       exn

  (rmt:send-receive 'no-sync-get/default #f `(,var ,default)))

(define (rmt:no-sync-del! var)
  (rmt:send-receive 'no-sync-del! #f `(,var)))

(define (rmt:no-sync-get-lock keyname)
  (rmt:send-receive 'no-sync-get-lock #f `(,keyname)))

(define (rmt:no-sync-add-job host-type vars-list exekey cmdline)
  (rmt:send-receive 'no-sync-add-job #f `(,host-type ,vars-list ,exekey ,cmdline)))

(define (rmt:no-sync-take-job host-type)
  (rmt:send-receive 'no-sync-take-job #f `(,host-type)))

(define (rmt:no-sync-job-records-clean)
  (rmt:set-receive 'no-sync-job-records-clean #f '()))

;;======================================================================
;; A R C H I V E S
;;======================================================================

(define (rmt:archive-get-allocations  testname itempath dneeded)
  (rmt:send-receive 'archive-get-allocations #f (list testname itempath dneeded)))

     ;; If no resources are available just kill time and loop again
     ;;
     ((not have-resources) ;; simply try again after waiting a second
      (if (runs:lownoise "no resources" 60)
	  (debug:print-info 1 *default-log-port* "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) ;; changed back to 1 from 0.25
      ;; could have done hed tal here but doing car/cdr of newtal to rotate tests
      (list (car newtal)(cdr newtal) reg reruns))
     
     ;; This is the final stage, everything is in place so launch the test
     ;;
     ((and have-resources
	   (or (null? prereqs-not-met)