Megatest

(declare (unit client))

(declare (uses common))
(declare (uses db))
(declare (uses tasks)) ;; tasks are where stuff is maintained about what is running.

(module client
*

)

(import client)

(include "common_records.scm")
(include "db_records.scm")

;; client:get-signature
(define (client:get-signature)
  (if *my-client-signature* *my-client-signature*
      (let ((sig (conc (get-host-name) " " (current-process-id))))
	(set! *my-client-signature* sig)
	*my-client-signature*)))

;; Not currently used! But, I think it *should* be used!!!
#;(define (client:logout serverdat)
  (let ((ok (and (socket? serverdat)
		 (cdb:logout serverdat *toppath* (client:get-signature)))))
    ok))




;; Do all the connection work, look up the transport type and set up the
;; connection if required.
;;
;; There are two scenarios. 
;;   1. We are a test manager and we received *transport-type* and *runremote* via cmdline
;;   2. We are a run tests, list runs or other interactive process and we must figure out
;;      *transport-type* and *runremote* from the monitor.db
;;
;; client:setup
;;
;; lookup_server, need to remove *runremote* stuff
;;
 
;;(define (http-transport:server-dat-make-url runremote)
(define (client:get-url runremote)
  (if (and (remote-iface runremote)
	   (remote-port  runremote))
      (conc "http://" 
	    (remote-iface runremote)
	    ":"
	    (remote-port  runremote))
      #f))

(define (client:setup areapath runremote #!key (remaining-tries 100) (failed-connects 0)(area-dat #f))
  (mutex-lock! *rmt-mutex*)
  (let ((res (client:setup-http areapath runremote remaining-tries: remaining-tries failed-connects: failed-connects area-dat: area-dat)))
    (mutex-unlock! *rmt-mutex*)
    res))

(define (client:setup-http areapath runremote #!key (remaining-tries 100) (failed-connects 0))
  (debug:print-info 2 *default-log-port* "client:setup remaining-tries=" remaining-tries)
  (server:start-and-wait areapath)
  (if (<= remaining-tries 0)
      (begin
	(debug:print-error 0 *default-log-port* "failed to start or connect to server")
	(exit 1))
      ;;
      ;; Alternatively here, we can get the list of candidate servers and work our way
      ;; through them searching for a good one.
      ;;
      (let* ((server-dat (server:choose-server areapath 'best))) ;; list host port start-time server-id pid
;;	     (runremote  (or area-dat *runremote*)))
	(if (not server-dat) ;; no server found
	    (begin
	      (if (< remaining-tries 99)(thread-sleep! 1)) ;; obviously it needs time
	      (client:setup-http areapath runremote remaining-tries: (- remaining-tries 1)))
	    (match server-dat
	      ((host port start-time server-id pid)
	       (debug:print-info 4 *default-log-port* "client:setup server-dat=" server-dat ", remaining-tries=" remaining-tries)

	       (if (not runremote)
                   (begin
		     ;; Here we are creating a runremote where there was none or it was clobbered with #f
		     ;;
		     (set! runremote (make-remote))
                     (let* ((server-info (server:check-if-running areapath)))
		       (remote-server-info-set! runremote server-info)
                       (if server-info
                           (begin
                             (remote-server-url-set! runremote (server:record->url server-info))
                             (remote-server-id-set! runremote (server:record->id server-info)))))))
	       ;; at this point we have a runremote
	       (if (and host port server-id)
		   (let* ((nada     (client:connect host port server-id runremote))
			  (ping-res (rmt:login-no-auto-client-setup runremote)))

		     (if ping-res

			 (if runremote
			     (begin

			       (debug:print-info 2 *default-log-port* "connected to " (client:get-url runremote))
			       runremote)
			     (client:setup-http areapath runremote remaining-tries: (- remaining-tries 1)))
			 (begin    ;; login failed but have a server record, clean out the record and try again
			   (debug:print-info 0 *default-log-port* "client:setup, login unsuccessful, will attempt to start server ... ping-res=" ping-res ", server-dat=" server-dat) ;; had runid.  Fixes part of Randy;s ticket 1405717332

			   (http-transport:close-connections runremote)



			   (thread-sleep! 1)
			   (client:setup-http areapath runremote remaining-tries: (- remaining-tries 1))
			   )))
		   (begin    ;; no server registered
		     ;; (server:kind-run areapath)
		     (server:start-and-wait areapath)
		     (debug:print-info 0 *default-log-port* "client:setup, no server registered, remaining-tries=" remaining-tries)
		     (thread-sleep! 1) ;; (+ 5 (random (- 20 remaining-tries))))  ;; give server a little time to start up, randomize a little to avoid start storms.
		     (client:setup-http areapath runremote remaining-tries: (- remaining-tries 1)))))
	      (else
	       (debug:print 0 *default-log-port* "ERROR: malformed server-dat="server-dat)))))))

;;
;; connect - stored in remote-condat
;;
;; (define (http-transport:client-connect iface port server-id runremote)
(define (client:connect iface port server-id runremote-in)
  (let* ((runremote (or runremote-in
			(make-runremote))))
    (debug:print-info 2 *default-log-port* "Connecting to server at "iface":"port", id "server-id)
    (let* ((api-url      (conc "http://" iface ":" port "/api"))
	   (api-uri      (uri-reference (conc "http://" iface ":" port "/api")))
	   (api-req      (make-request method: 'POST uri: api-uri)))
      ;;	 (server-dat   (vector iface port api-uri api-url api-req (current-seconds) server-id)))
      (remote-iface-set!   runremote iface)
      (remote-port-set!    runremote port)
      (remote-server-id-set! runremote server-id)
      (remote-connect-time-set! runremote (current-seconds))
      (remote-last-access-set! runremote (current-seconds))
      (remote-api-url-set! runremote api-url)
      (remote-api-uri-set! runremote api-uri)
      (remote-api-req-set! runremote api-req)
      runremote)))

    ((abort) "ABORT")
    ((skip) "SKIP")
    (else "FAIL")))

(define (common:logpro-exit-code->test-status exit-code)
  (status-sym->string (common:logpro-exit-code->status-sym exit-code)))

;; 
(defstruct remote
  (hh-dat            (let ((res (or (server:choose-server *toppath* 'homehost)
				    (cons #f #f))))
		       (assert (pair? res)(conc "FATAL: hh-dat should be a pair, got "res))
		       res))
  (server-url        #f) ;; (server:check-if-running *toppath*) #f))
  (server-id         #f)
  (server-info       #f) ;; (if *toppath* (server:check-if-running *toppath*) #f))
  (last-server-check 0)  ;; last time we checked to see if the server was alive
  (connect-time      (current-seconds)) ;; when we first connected
  (last-access       (current-seconds)) ;; last time we talked to server
  (conndat           #f) ;; iface port api-uri api-url api-req seconds server-id

  (server-timeout    (server:expiration-timeout))
  (force-server      #f)
  (ro-mode           #f)  
  (ro-mode-checked   #f) ;; flag that indicates we have checked for ro-mode

  ;; conndat stuff
  (iface             #f) ;; TODO: Consolidate this data with server-url and server-info above
  (port              #f)
  (api-url           #f)
  (api-uri           #f)
  (api-req           #f))

;; launching and hosts
(defstruct host
  (reachable    #f)
  (last-update  0)
  (last-used    0)
  (last-cpuload 1))

(define (common:version-db-delta)
  (- megatest-version (common:get-last-run-version-number)))

(define (common:version-changed?)
  (not (equal? (common:get-last-run-version)
               (common:version-signature))))


;; From 1.70 to 1.80, db's are compatible.

(define (common:api-changed?)
  (let* (
    (megatest-major-version (substring (->string megatest-version) 0 4))
    (run-major-version (substring (conc (common:get-last-run-version)) 0 4))
   )
   (and (not (equal? megatest-major-version "1.80"))
     (not (equal? megatest-major-version megatest-run-version)))
  )
)

;;======================================================================
;; Move me elsewhere ...
;; RADT => Why do we meed the version check here, this is called only if version misma
;;
(define (common:cleanup-db dbstruct #!key (full #f))
  (apply db:multi-db-sync 

;;          logs directory you wish to log-rotate.
;;
(define (common:rotate-logs)
  (let* ((all-files (make-hash-table))
	 (stats     (make-hash-table))
	 (inc-stat  (lambda (key)
		      (hash-table-set! stats key (+ (hash-table-ref/default stats key 0) 1))))
	(max-allowed (string->number (or (configf:lookup *configdat* "setup" "max-logfiles") "600")))) ;; name -> age
    (if (not (directory-exists? "logs"))(create-directory "logs"))
    (directory-fold 
     (lambda (file rem)
       (handle-exceptions
	exn
	(begin
	  (debug:print-info 2 *default-log-port* "unable to rotate log " file ", probably handled by another process, this is safe to ignore. exn=" exn)

	    (debug:print-info 0 *default-log-port* "Deleted " (length files) " files from logs, keeping " max-allowed " files."))))))

;;======================================================================
;; Force a megatest cleanup-db if version is changed and skip-version-check not specified
;; Do NOT check if not on homehost!
;;
(define (common:exit-on-version-changed)
  (if (and *toppath*              ;; do nothing if *toppath* not yet provided
	   (common:on-homehost?))
      (if (common:api-changed?)
	  (let* ((mtconf (conc (get-environment-variable "MT_RUN_AREA_HOME") "/megatest.config"))
                 (dbfile  (conc (get-environment-variable "MT_RUN_AREA_HOME") ".megatest/main.db"))
                 (read-only (not (file-write-access? dbfile)))
                 (dbstruct (db:setup #t))) ;; (db:setup-db *dbstruct-dbs* *toppath* #f))) ;;  #t)))
	    (debug:print 0 *default-log-port*
			 "WARNING: Version mismatch!\n"
			 "   expected: " (common:version-signature) "\n"
			 "   got:      " (common:get-last-run-version))
            (cond

                 (print-call-chain (current-error-port))
                 (exit 1))
               (common:cleanup-db dbstruct)))
             ((not (common:file-exists? mtconf))
              (debug:print 0 *default-log-port* "   megatest.config does not exist in this area.  Cannot proceed with megatest version migration.")
              (exit 1))
             ((not (common:file-exists? dbfile))
              (debug:print 0 *default-log-port* "   .megatest/main.db does not exist in this area.  Cannot proceed with megatest version migration.")
              (exit 1))
             ((not (eq? (current-user-id)(file-owner mtconf)))
              (debug:print 0 *default-log-port* "   You do not own .megatest/main.db in this area.  Cannot proceed with megatest version migration.")
              (exit 1))
             (read-only
              (debug:print 0 *default-log-port* "   You have read-only access to this area.  Cannot proceed with megatest version migration.")
              (exit 1))
             (else
              (debug:print 0 *default-log-port* " to switch versions you can run: \"megatest -cleanup-db\"")
              (exit 1)))))))

    (if dat
	dat
	""))))

(define (common:alist-ref/default key alist default)
  (or (alist-ref key alist) default))

;; moved into commonmod
;;
;; (define (common:low-noise-print waitval . keys)
;;   (let* ((key      (string-intersperse (map conc keys) "-" ))
;; 	 (lasttime (hash-table-ref/default *common:denoise* key 0))
;; 	 (currtime (current-seconds)))
;;     (if (> (- currtime lasttime) waitval)
;; 	(begin
;; 	  (hash-table-set! *common:denoise* key currtime)
;; 	  #t)
;; 	#f)))

(define (common:get-megatest-exe)
  (or (getenv "MT_MEGATEST") "megatest"))

(define (common:read-encoded-string instr)
  (handle-exceptions
   exn

      *db-cache-path*
      (if *toppath* ;; common:get-create-writeable-dir
	  (handle-exceptions
	      exn
	      (begin
		(debug:print-error 0 *default-log-port* "Couldn't create path to " *db-cache-path* ", exn=" exn)
		(exit 1))
	      (let* ((toppath (common:real-path *toppath*))
		     (tsname (common:get-testsuite-name))
		     (dbpath (common:get-create-writeable-dir
			      (list (conc "/tmp/" (current-user-name)
					  "/megatest_localdb/"
					  tsname "/"
					  (string-translate toppath "/" "."))
				    (conc "/tmp/" (current-process-id) ;; just in case we have an issue with the dir by own user name
					  "/megatest_localdb/"
					  tsname
					  (string-translate toppath "/" "."))
				    ))))
		(set! *db-cache-path* dbpath)
		;; ensure megatest area has .megatest
		(let ((dbarea (conc *toppath* "/.megatest")))
		  (if (not (file-exists? dbarea))
		      (create-directory dbarea)))
		;; ensure tmp area has .megatest

  (message-digest-string (md5-primitive) str))

;;======================================================================
;; E X I T   H A N D L I N G
;;======================================================================

(define (common:run-sync?)
  (and *toppath*               ;; gate if called before *toppath* is set
       (common:on-homehost?)
       (args:get-arg "-server")))

(define (common:human-time)
  (time->string (seconds->local-time (current-seconds)) "%Y-%m-%d %H:%M:%S"))


(define (std-signal-handler signum)
  ;; (signal-mask! signum)

      (begin
	(debug:print-error 0 *default-log-port* "command \"/bin/readlink -f " path "\" failed. exn=" exn)
	path) ;; just give up
    (with-input-from-pipe
	(conc "/bin/readlink -f " path)
      (lambda ()
	(read-line)))))

;; for reasons I don't understand multiple calls to real-path in parallel threads
;; must be protected by mutexes
;;
(define (common:real-path inpath)
  ;; (process:cmd-run-with-stderr->list "readlink" "-f" inpath)) ;; cmd . params)
  ;; (let-values 
  ;;  (((inp oup pid) (process "readlink" (list "-f" inpath))))
  ;;  (with-input-from-port inp
  ;;    (let loop ((inl (read-line))
  ;;       	(res #f))
  ;;      (print "inl=" inl)
  ;;      (if (eof-object? inl)
  ;;          (begin
  ;;            (close-input-port inp)
  ;;            (close-output-port oup)
  ;;            ;; (process-wait pid)
  ;;            res)
  ;;          (loop (read-line) inl))))))
  (with-input-from-pipe (conc "readlink -f " inpath) read-line))

;;======================================================================
;; returns *effective load* (not normalized)
;;
(define (common:get-intercept onemin fivemin)
  (if (< onemin fivemin) ;; load is decreasing, just use the onemin load
      onemin

		  (begin ;; found a host, return it
		    (debug:print 0 *default-log-port* "INFO: Found host: " new-best " load: " load " last-used: " delta " seconds ago, with job-rate: " job-rate)
		    (host-last-used-set! rec curr-time)
		    new-best)
		  (if (null? tal) #f (loop (car tal)(cdr tal) best-host)))))))))

(define (common:wait-for-homehost-load maxnormload msg)
  (let loop ((start-time (current-seconds))) ;; we saw some instances of this being called before *toppath* was set. This might be an early setup race. This delay should help but it is impossible to test...
    (if (not *toppath*)
	(begin
	  (debug:print 0 *default-log-port* "ERROR: common:wait-for-homehost-load called before *toppath* set.")
	  (thread-sleep! 30)
	  (if (< (- (current-seconds) start-time) 300)
	      (loop start-time)))))
  (let* ((hh-dat (if (common:on-homehost?) ;; if we are on the homehost then pass in #f so the calls are local.
                     #f
                     (server:choose-server *toppath* 'homehost)))
         (hh     (if hh-dat (car hh-dat) #f)))
    (common:wait-for-normalized-load maxnormload msg hh)))

(define (common:get-num-cpus remote-host)

(define (get-uname . params)
  (let* ((uname-res (process:cmd-run->list (conc "uname " (if (null? params) "-a" (car params)))))
	 (uname #f))
    (if (null? (car uname-res))
	"unknown"
	(caar uname-res))))





















;;======================================================================
;; D I S K   S P A C E 
;;======================================================================

(define (common:get-disk-space-used fpath)
  (with-input-from-pipe (conc "/usr/bin/du -s " fpath) read))

;; return the target db handle so it can be used
;;
(define (db:cache-for-read-only source target #!key (use-last-update #f))
  (if (and (hash-table-ref/default *global-db-store* target #f)
	   (>= (file-modification-time target)(file-modification-time source)))
      (hash-table-ref *global-db-store* target)
      (let* ((toppath   (launch:setup))
	     (targ-db-last-mod (db:get-sqlite3-mod-time target))
;;	      (if (common:file-exists? target)
;; BUG: This needs to include wal mode stuff .shm etc.
;;				   (file-modification-time target)
;;				   0))
	     (cache-db  (or (hash-table-ref/default *global-db-store* target #f)
			    (db:open-megatest-db path: target)))
	     (source-db (db:open-megatest-db path: source))
	     (curr-time (current-seconds))
	     (res      '())
	     (last-update (if use-last-update (cons "last_update" targ-db-last-mod) #f)))
	(db:sync-tables (db:sync-main-list source-db) last-update source-db cache-db)

;; 			  megatest-db
;; 			  (conc cache-dir "/" fname)
;; 			  use-last-update: #t)))
;; 	  (thread-start! th1)
;; 	  (apply proc cache-db params)
;; 	  ))))

(define (db:get-sqlite3-mod-time fname)
  (let* ((wal-file (conc fname "-wal"))
	 (shm-file (conc fname "-shm"))
	 (get-mtime (lambda (f)
		      (if (and (file-exists? f)
			       (file-read-access? f))
			  (file-modification-time f)
			  0))))
    (max (get-mtime fname)
	 (get-mtime wal-file)
	 (get-mtime shm-file))))
	 
(define (db:all-db-sync dbstruct)
  (let* ((dbdat (db:open-db dbstruct #f db:initialize-main-db))
	 (data-synced       0) ;; count of changed records
    (tmp-area       (common:get-db-tmp-area))
    (dbfiles        (glob (conc tmp-area"/.megatest/*.db")))
    (sync-durations (make-hash-table))
    (no-sync-db        (db:open-no-sync-db)))
    (for-each
     (lambda (file) ;; tmp db file
       (debug:print-info 3 *default-log-port* "file: " file)
       (let* ((fname       (conc (pathname-file file) ".db")) ;; fname is tmp db file
              (wal-file (conc fname "-wal"))
              (shm-file (conc fname "-shm"))
	      (fulln       (conc *toppath*"/.megatest/"fname)) ;; fulln is nfs db name
              (wal-time     (if (file-exists? wal-file)             
			       (file-modification-time wal-file)
                               0))
              (shm-time     (if (file-exists? shm-file)             
			       (file-modification-time shm-file)
                               0))
	      (time1        (db:get-sqlite3-mod-time file))
;;	       (if (file-exists? file)              ;; time1 is the max itime of the tmp db, -wal and -shm files.
;;			       (max (file-modification-time file) wal-time shm-time)
;;			       (begin
;;				 (debug:print-info 2 *default-log-port* "Sync - I do not see file "file)
;;			   1)))
	      (time2       (db:get-sqlite3-mod-time fulln))
;;	      (if (file-exists? fulln)             ;; time2 is nfs file time
;;			       (file-modification-time fulln)
;;			       (begin
;;				 (debug:print-info 2 *default-log-port* "Sync - I do not see file "fulln)
;;				 0)))
	      (changed      (> (- time1 time2) (+ (random 5) 1)))  ;; it has been at some few seconds since last synced
	      (changed10    (> (- time1 time2) 10)) ;; it has been at least ten seconds since sync'd
	      (jfile-exists (file-exists? (conc file"-journal"))) ;; i.e. are we busy?
	      (do-cp        (cond
			     ((not (file-exists? fulln)) ;; shouldn't happen, but this might recover
			      (cons #t (conc "File "fulln" not found! Copying "fname" to "fulln)))
			     ((and (not jfile-exists) changed)

    (servers (server:choose-server *toppath* 'all-valid)) ;; (server:get-list *toppath*))
    (src-area (if old2new *toppath* tmp-area))
    (dest-area (if old2new tmp-area *toppath*))
    (dbfiles        (if old2new (glob (conc *toppath* "/.megatest/*.db")) (glob (conc tmp-area "/.megatest/*.db"))))
    (keys (db:get-keys dbstruct))
    (sync-durations (make-hash-table)))


    (if (and killservers servers)
      (begin
       	  (for-each
	   (lambda (server)
             (handle-exceptions
             exn
             (begin 
               (debug:print-info 0 *default-log-port*  "Unable to get host and/or port from " server ", exn=" exn)     
               #f)
	     (match-let (((mod-time host port start-time server-id pid) server))
	       (if (and host pid)
		   (tasks:kill-server host pid)))))
	   servers)
          (delete-file* (common:get-sync-lock-filepath))
      )
    )

    (if (not dbfiles)
      (debug:print-error 0 *default-log-port* "no dbfiles found in " (conc *toppath* "/.megatest"))
    (for-each
     (lambda (srcfile)
       (debug:print-info 3 *default-log-port* "file: " srcfile)
       (let* ((fname (conc (pathname-file srcfile) ".db"))
              (basename (pathname-file srcfile))
              (run-id (if (string= basename "main") #f (string->number basename)))
	      (destfile (conc dest-area "/.megatest/" fname))

			   (debug:print-info 0 *default-log-port* "Sync - I do not see file " destfile)
			   0)))
	      (changed ( < (- time2 time1) 6.0)) ;; dest db not updated within last 6 seconds

      (do-cp (cond
		      ((not (file-exists? destfile)) ;; shouldn't happen, but this might recover
		       (debug:print-info 2 *default-log-port* "File " destfile " not found. Copying "srcfile" to "destfile)
                       ;; TODO: Need to fix this for WAL mod. Can't just copy.
                       (system (conc "/bin/mkdir -p " dest-directory))
                       (system (conc "/bin/cp " srcfile " " destfile))
		       #t)
		      (changed ;; (and changed
		       ;; (> (- (current-seconds) time1) 3)) ;; if file is changed and three seconds have passed.
		       #t)
		      ((and changed *time-to-exit*) ;; last sync

               )
	       (hash-table-set! sync-durations (conc srcfile ".db") (- (current-milliseconds) start-time)))
	     (debug:print-info 2 *default-log-port* "skipping delta sync. " srcfile " is up to date")
          )
       )
     )
     dbfiles
    )
    )
    data-synced
  )
)



;; Sync all changed db's
;;
(define (db:tmp->megatest.db-sync dbstruct run-id last-update)
  (let* ((subdbs (hash-table-values (dbr:dbstruct-subdbs dbstruct)))
	 (res    '()))
    (for-each

  (db:with-db
   dbstruct #f #f
   (lambda (dbdat db)
       (if msg
         (sqlite3:execute db "UPDATE runs SET status=?,comment=? WHERE id=?;" status msg run-id)
         (sqlite3:execute db "UPDATE runs SET status=? WHERE id=?;" status run-id)))))

(define (db:set-run-state-status-db db run-id state status )
  (sqlite3:execute db "UPDATE runs SET status=?,state=? WHERE id=?;" status state run-id))

(define (db:set-run-state-status dbstruct run-id state status )
  (db:with-db
   dbstruct #f #f
   (lambda (dbdat db)

     (db:set-run-state-status-db db run-id state status))))
     

(define (db:get-run-status dbstruct run-id)
  (let ((res "n/a"))
    (db:with-db
     dbstruct #f #f
     (lambda (dbdat db)
       (sqlite3:for-each-row 
	(lambda (status)

    (db:with-db dbstruct run-id #f
		(lambda (dbdat db)
		  (sqlite3:for-each-row
		   (lambda (run-id testname item-path state status)
		     ;; id,run_id,testname,state,status,event_time,host,cpuload,diskfree,uname,rundir,item_path,run_duration,final_logf,comment
		     (set! res (vector test-id run-id testname state status -1 "" -1 -1 "" "-" item-path -1 "-" "-")))
		   db 
		   "SELECT run_id,testname,item_path,state,status FROM tests WHERE id=? and run_id=?;" 
		   test-id run-id)))
    res))

;; get a useful subset of the tests data (used in dashboard
;; use db:mintest-get-{id ,run_id,testname ...}
;;
(define (db:get-tests-for-run-mindata dbstruct run-id testpatt states statuses not-in)
  (db:get-tests-for-run dbstruct run-id testpatt states statuses #f #f not-in #f #f "id,run_id,testname,state,status,event_time,item_path" 0 #f))

;; ;; NB// Ultimately this will be deprecated in deference to mt:test-set-state-status-by-id
;;
;;      NOTE: run-id is not used
;; ;;
(define (db:test-set-state-status dbstruct run-id test-id newstate newstatus newcomment)
  (db:with-db
   dbstruct
   run-id #f

   (lambda (dbdat db)
     (db:test-set-state-status-db db run-id test-id newstate newstatus newcomment))))

(define (db:test-set-state-status-db db run-id test-id newstate newstatus newcomment)
  (cond
   ((and newstate newstatus newcomment)
    (sqlite3:execute db "UPDATE tests SET state=?,status=?,comment=? WHERE id=?;" newstate newstatus newcomment ;; (sdb:qry 'getid newcomment)
		     test-id))
   ((and newstate newstatus)
    (sqlite3:execute db "UPDATE tests SET state=?,status=? WHERE id=?;" newstate newstatus test-id))
   (else
    (if newstate   (sqlite3:execute db "UPDATE tests SET state=?   WHERE id=?;" newstate   test-id))
    (if newstatus  (sqlite3:execute db "UPDATE tests SET status=?  WHERE id=?;" newstatus  test-id))
    (if newcomment (sqlite3:execute db "UPDATE tests SET comment=? WHERE id=?;" newcomment ;; (sdb:qry 'getid newcomment)
				    test-id))))
  ;; (mt:process-triggers dbstruct run-id test-id newstate newstatus)) ;; NOTE: Moved into calling function
  )

;; NEW BEHAVIOR: Count tests running in all runs!
;;
(define (db:get-count-tests-running dbstruct run-id) ;; fastmode)
  (let* ((qry ;; (if fastmode
		;;   "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '') LIMIT 1;"
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '');")) ;; )

(define (db:get-test-info dbstruct run-id test-name item-path)
  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (dbdat db)
     (db:get-test-info-db db run-id test-name item-path))))

(define (db:get-test-info-db db run-id test-name item-path)
  (let ((res #f))
    (sqlite3:for-each-row
     (lambda (a . b)
       (set! res (apply vector a b)))
     db
     (conc "SELECT " db:test-record-qry-selector " FROM tests WHERE testname=? AND item_path=? AND run_id=?;")
     test-name item-path run-id)
    res))

(define (db:test-get-rundir-from-test-id dbstruct run-id test-id)
  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (dbdat db)

	 (test-name    (if (number? test-name)
			   (db:test-get-testname testdat)
			   test-name))
	 (item-path    (db:test-get-item-path testdat))
         (tl-testdat   (db:get-test-info dbstruct run-id test-name ""))
         (tl-test-id   (if tl-testdat
			   (db:test-get-id tl-testdat)
			   #f))
	 (new-state-eh #f)
	 (new-status-eh #f))
    (if (member state '("LAUNCHED" "REMOTEHOSTSTART")) 
	(db:general-call dbstruct run-id 'set-test-start-time (list test-id)))
    (mutex-lock! *db-transaction-mutex*)
    (db:with-db
     dbstruct run-id #f
     (lambda (dbdat db)
       (let ((tr-res
              (sqlite3:with-transaction
               db
               (lambda ()
                 ;; NB// Pass the db so it is part fo the transaction
                 (db:test-set-state-status-db db run-id test-id state status comment) ;; this call sets the item state/status
                 (if (not (equal? item-path "")) ;; only roll up IF incoming test is an item
                     (let* ((state-status-counts (db:get-all-state-status-counts-for-test db run-id test-name item-path state status)) ;; item-path is used to exclude current state/status of THIS test
			    (state-statuses      (db:roll-up-rules state-status-counts state status))
                            (newstate            (car state-statuses))
                            (newstatus           (cadr state-statuses)))
		       (set! new-state-eh newstate)
		       (set! new-status-eh newstatus)
                       (debug:print 4 *default-log-port* "BB> tl-test-id="tl-test-id" ; "test-name":"item-path" newstate="newstate" newstatus="newstatus" len(sscs)="(length state-status-counts)  " state-status-counts: "
				    (apply conc
					   (map (lambda (x)
						  (conc
                     				   (with-output-to-string (lambda () (pp (dbr:counts->alist x)))) " | "))
						state-status-counts))); end debug:print

		       (if tl-test-id
			   (db:test-set-state-status-db db run-id tl-test-id newstate newstatus #f)) ;; we are still in the transaction - must access the db and not the dbstruct
		       ))))))
         (mutex-unlock! *db-transaction-mutex*)
         (if (and test-id state status (equal? status "AUTO")) 
             (db:test-data-rollup dbstruct run-id test-id status))
	 (if new-state-eh ;; moved from db:test-set-state-status
	      (mt:process-triggers dbstruct run-id test-id new-state-eh new-status-eh))
         tr-res)))))

(define (db:roll-up-rules state-status-counts state status)
  (let* ((running     (length (filter (lambda (x)
					(member (dbr:counts-state x) *common:running-states*))
				      state-status-counts)))
	 (bad-not-started      (length (filter (lambda (x)

    (db:with-db
     dbstruct run-id #f
     (lambda (dbdat db)
       (let ((tr-res
              (sqlite3:with-transaction
               db
               (lambda ()
                   (let* ((state-status-counts (db:get-all-state-status-counts-for-run-db db run-id))
			  (state-statuses      (db:roll-up-rules state-status-counts #f #f ))
                          (newstate            (car state-statuses))
                          (newstatus           (cadr state-statuses))) 
		     (if (or (not (eq? newstate curr-state)) (not (eq?  newstatus curr-status)))
			 (db:set-run-state-status-db db run-id newstate newstatus )))))))
         (mutex-unlock! *db-transaction-mutex*)
         tr-res))))

(define (db:get-all-state-status-counts-for-run-db db run-id)



  (sqlite3:map-row
   (lambda (state status count)
     (make-dbr:counts state: state status: status count: count))
   db
   "SELECT state,status,count(id) FROM tests WHERE run_id=?  GROUP BY state,status;"
   run-id ))


(define (db:get-all-state-status-counts-for-run dbstruct run-id)
  (db:with-db
   dbstruct #f #f
   (lambda (dbdat db)
     (db:get-all-state-status-counts-for-run-db dbstruct run-id))))

;; BBnote: db:get-all-state-status-counts-for-test returns dbr:counts object aggregating state and status of items of a given test, *not including rollup state/status*
;;
;; NOTE: This is called within a transaction
;;
(define (db:get-all-state-status-counts-for-test db run-id test-name item-path item-state-in item-status-in)
  (let* ((test-info   (db:get-test-info-db db run-id test-name item-path))
         (item-state  (or item-state-in (db:test-get-state test-info))) 
         (item-status (or item-status-in (db:test-get-status test-info)))
         (other-items-count-recs (sqlite3:map-row



                                  (lambda (state status count)
                                    (make-dbr:counts state: state status: status count: count))
                                  db
                                  ;; ignore current item because we have changed its value in the current transation so this select will see the old value.
                                  "SELECT state,status,count(id) FROM tests WHERE run_id=? AND testname=? AND item_path != '' AND item_path !=? GROUP BY state,status;"
                                  run-id test-name item-path))

         ;; add current item to tally outside of sql query
	 (match-countrec-lambda (lambda (countrec) 
				  (and (equal? (dbr:counts-state  countrec) item-state)
                                       (equal? (dbr:counts-status countrec) item-status))))
	 
         (already-have-count-rec-list
          (filter match-countrec-lambda other-items-count-recs)) ;; will have either 0 or 1 count recs depending if another item shares this item's state/status
         
         (updated-count-rec    (if (null? already-have-count-rec-list)
                                   (make-dbr:counts state: item-state status: item-status count: 1)
                                   (let* ((our-count-rec (car already-have-count-rec-list))
                                          (new-count (add1 (dbr:counts-count our-count-rec))))
                                     (make-dbr:counts state: item-state status: item-status count: new-count))))

         (nonmatch-countrec-lambda (lambda (countrec) (not (match-countrec-lambda countrec))))
         
         (unrelated-rec-list   
          (filter nonmatch-countrec-lambda other-items-count-recs)))

    (cons updated-count-rec unrelated-rec-list)))

;; (define (db:get-all-item-states db run-id test-name)
;;   (sqlite3:map-row 
;;    (lambda (a) a)
;;    db
;;    "SELECT DISTINCT state FROM tests WHERE item_path != '' AND state != 'DELETED' AND run_id=? AND testname=?"

					  (sqlite3:finalize! db #t)
					  ;; (vector-set! *task-db* 0 #f)
					  (set! *task-db* #f)))))
			      (if (and (not (args:get-arg "-server"))
				       *runremote*)
				  (begin
				    (debug:print-info 0 *default-log-port* "Closing all client connections...")
				    (http-transport:close-connections *runremote*)
				    #;(http-client#close-all-connections!)))
                              ;; (if (and *runremote*
                              ;;          (remote-conndat *runremote*))
                              ;;     (begin
                              ;;       (http-client#close-all-connections!))) ;; for http-client
                              (if (not (eq? *default-log-port* (current-error-port)))
                                  (close-output-port *default-log-port*))
			      (set! *default-log-port* (current-error-port))) "Cleanup db exit thread"))

(define (dbfile:print-err . params)
  (with-output-to-port
      (current-error-port)
    (lambda ()
      (apply print params))))
    
(define (dbfile:cautious-open-database fname init-proc sync-mode journal-mode #!optional (tries-left 500))
  (let* ((busy-file  (conc fname "-journal"))
	 (delay-time (* (- 51 tries-left) 1.1))
      	 (write-access (file-write-access? fname))
         (dir-access (file-write-access? (pathname-directory fname)))
         (retry      (lambda ()
		       (thread-sleep! delay-time)
		       (if (> tries-left 0)
			   (dbfile:cautious-open-database fname init-proc
							  sync-mode: sync-mode journal-mode
							  (- tries-left 1))))))
    (assert (>= tries-left 0) (conc "FATAL: too many attempts in dbfile:cautious-open-database of "fname", giving up."))
    
    (if (and (file-write-access? fname)
	     (file-exists? busy-file))
	(begin
	  (if (common:low-noise-print 120 busy-file)
	      (dbfile:print-err "INFO: dbfile:cautious-open-database: journal file "
				busy-file" exists, trying again in few seconds."))
	  (thread-sleep! 1)
	  (if (eq? tries-left 2)
	      (begin
	  	(dbfile:print-err "INFO: forcing journal rollup "busy-file)
	  	(dbfile:brute-force-salvage-db fname)))
	  (dbfile:cautious-open-database fname init-proc sync-mode journal-mode (- tries-left 1)))
	
	(let* ((result (condition-case
		        (if dir-access
			    (dbfile:with-simple-file-lock
			     (conc fname ".lock")
			     (lambda ()
			       (let* ((db-exists (file-exists? fname))

	  (dbfile:print-err "INFO: db:lock-and-delta-sync copying db "runid" at "(current-seconds))
	  (set! *db-sync-in-progress* #t)
	  (db:sync-touched dbstruct runid keys dbinit)
	  (set! *db-sync-in-progress* #f)
	  (delete-file* lock-file)
	  #t)
        (begin
	  (if (common:low-noise-print 120 (conc "no lock "from-db-file))
	      (dbfile:print-err "INFO: could not get lock for " from-db-file ", sync likely in progress."))
	  #f
	  ))))

;; ;; Get a lock from the no-sync-db for the from-db, then delta sync the from-db to the to-db, otherwise return #f
;; ;;
;; (define (db:lock-and-delta-sync-orig no-sync-db dbstruct from-db-file runid keys dbinit)
;;   (assert (not *db-sync-in-progress*) "FATAL: db:lock-and-sync called while a sync is in progress.")

	   '("owner"          #f)
	   '("description"    #f)
	   '("reviewed"       #f)
	   '("iterated"       #f)
	   '("avg_runtime"    #f)
	   '("avg_disk"       #f)
	   '("tags"           #f)
	   '("jobgroup"       #f))


     (list "tasks_queue"
           '("id"            #f)
           '("action"        #f)
           '("owner"         #f) 
           '("state"         #f)
           '("target"        #f)
           '("name"          #f)
           '("testpatt"      #f)
           '("keylock"       #f)
           '("params"        #f)
           '("creation_time" #f)
           '("execution_time" #f))
     )))

(define (db:sync-all-tables-list dbstruct keys)
  (append (db:sync-main-list dbstruct keys)
	  db:sync-tests-only))

;; tbls is ( ("tablename" ( "field1" [#f|proc1] ) ( "field2" [#f|proc2] ) .... ) )
;; db's are dbdat's

(define dbfile:db-init-proc (make-parameter #f))

;; (db:with-db dbstruct run-id sqlite3:exec "select blah fgrom blaz;")
;; r/w is a flag to indicate if the db is modified by this query #t = yes, #f = no
;;
(define (db:with-db dbstruct run-id r/w proc . params)
  (assert dbstruct "FATAL: db:with-db called with dbstruct "#f)
  (assert (dbr:dbstruct? dbstruct) "FATAL: dbstruct is "dbstruct)
  (let* ((use-mutex (> *api-process-request-count* 25)) ;; risk of db corruption
	 (have-struct (dbr:dbstruct? dbstruct))
         (dbdat     (if have-struct                ;; this stuff just allows us to call with a db handle directly
			(db:open-db dbstruct run-id (dbfile:db-init-proc)) ;; (dbfile:get-subdb dbstruct run-id)
			#f))
	 (db        (if have-struct                ;; this stuff just allows us to call with a db handle directly
			(dbr:dbdat-dbh dbdat)
			dbstruct))
	 (fname     (if dbdat
			(dbr:dbdat-dbfile dbdat)
			"nofilenameavailable"))
	 (jfile     (conc fname"-journal"))
	 #;(subdb     (if have-struct
			(dbfile:get-subdb dbstruct run-id)
			#f))
	 ) ;; was 25
    (assert (sqlite3:database? db) "FATAL: db:with-db, db is not a database, db="db", fname="fname)
    (if (file-exists? jfile)
	(begin
	  (dbfile:print-err "INFO: "jfile" exists, delaying to reduce database load")
	  (thread-sleep! 0.2)))
    (if (and use-mutex
	     (common:low-noise-print 120 "over-50-parallel-api-requests"))
	(dbfile:print-err *api-process-request-count* " parallel api requests being processed in process "

(import canvas-draw-iup)
(use regex typed-records matchable)

(declare (unit dcommon))

(declare (uses gutils))
(declare (uses db))
(declare (uses commonmod))

(import commonmod)
;; (declare (uses synchash))

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

	     (max-row      (if (null? row-indices) 1 (common:max (map cadr row-indices))))
	     (max-col      (if (null? col-indices) 1 
			       (common:max (map cadr col-indices))))
	     (max-visible  (max (- (dboard:tabdat-num-tests tabdat) 15) 3))
	     (max-col-vis  (if (> max-col 10) 10 max-col))
	     (numrows      1)
	     (numcols      1))
	(if (common:low-noise-print 60 "runs-stats-update-clear")
	    (iup:attribute-set! stats-matrix "CLEARVALUE" "CONTENTS"))
	(iup:attribute-set! stats-matrix "NUMCOL" max-col )
	(iup:attribute-set! stats-matrix "NUMLIN" (if (< max-row max-visible) max-visible max-row)) ;; min of 20
	(iup:attribute-set! stats-matrix "NUMCOL_VISIBLE" max-col-vis)
	(iup:attribute-set! stats-matrix "NUMLIN_VISIBLE" (if (> max-row max-visible) max-visible max-row))
	;;(print "row-indices: " row-indices " col-indices: " col-indices)
	;; Row labels
	(for-each (lambda (ind)

	(config-use-proxy (equal? (configf:lookup *configdat* "client" "use-http_proxy") "yes")))
    (if (not config-use-proxy)
	(determine-proxy (constantly #f)))
    (debug:print-info 0 *default-log-port* "http-transport:try-start-server time=" (seconds->time-string (current-seconds)) " ipaddrsstr=" ipaddrstr " portnum=" portnum " config-hostname=" config-hostname)
    (handle-exceptions
	exn
	(begin
	  ;; (print-error-message exn)
	  (if (< portnum 64000)
	      (begin 
		(debug:print 0 *default-log-port* "WARNING: attempt to start server failed. Trying again ...")
		(debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
		(debug:print 5 *default-log-port* "exn=" (condition->list exn))
		(portlogger:open-run-close portlogger:set-failed portnum)
		(debug:print 0 *default-log-port* "WARNING: failed to start on portnum: " portnum ", trying next port")

(define (http-transport:inc-requests-and-prep-to-close-all-connections)
  (mutex-lock! *http-mutex*)
  (set! *http-requests-in-progress* (+ 1 *http-requests-in-progress*)))

;; Send "cmd" with json payload "params" to serverdat and receive result
;;
(define (http-transport:client-api-send-receive run-id runremote cmd params #!key (numretries 3))



  (assert (remote? runremote) "FATAL: http-transport:client-api-send-receive called with serverdat="serverdat)

  (let* ((fullurl    (remote-api-req runremote))
	 (res        (vector #f "uninitialized"))
	 (success    #t)
	 (sparams    (db:obj->string params transport: 'http))

         (server-id  (remote-server-id runremote)))




       (debug:print-info 11 *default-log-port* "cmd=" cmd " fullurl=" fullurl " server-id=" server-id " current time:" (current-seconds)) 

       ;; set up the http-client here
       (max-retry-attempts 1)
       ;; consider all requests indempotent
       (retry-request? (lambda (request)
			 #f))

                                                    (debug:print-info 0 *default-log-port* "couldn't talk to server, trying again ...")
                                                    (begin
                                                      (debug:print 0 *default-log-port* "WARNING: failure in with-input-from-request to " fullurl ".")
                                                      (debug:print 0 *default-log-port* " message: " msg ", exn=" exn)
                                                      (debug:print 0 *default-log-port* " cmd: " cmd " params: " params " key:" (or server-id "thekey"))
                                                      (debug:print 0 *default-log-port* " call-chain: " call-chain)))
						;; what if another thread is communicating ok? Can't happen due to mutex


						(http-transport:close-connections runremote)



						(mutex-unlock! *http-mutex*)



						;; (close-connection! fullurl)
						(db:obj->string #f))
					      (with-input-from-request ;; was dat
					       fullurl 
					       (list (cons 'key (or server-id   "thekey"))
						     (cons 'cmd cmd)
						     (cons 'params sparams))
					       read-string))

	     (signal (make-composite-condition
		      (make-property-condition 
		       'timeout
		       'message "nmsg-transport:client-api-send-receive-raw timed out talking to server")))))))

;; careful closing of connections stored in *runremote*
;;
(define (http-transport:close-connections runremote)


  (if (remote? runremote)


      (let ((api-dat (remote-api-uri runremote)))
	(handle-exceptions
	    exn
	  (begin
	    (print-call-chain *default-log-port*)
	    (debug:print-error 0 *default-log-port* " closing connection failed with error: " ((condition-property-accessor 'exn 'message) exn) ", exn=" exn))
	  (if (args:any-defined? "-server" "-execute" "-run")
	      (debug:print-info 0 *default-log-port* "Closing connections to "api-dat))
	  (if api-dat (close-connection! api-dat))
	  (remote-conndat-set! runremote #f)
	  #t))




















      #f))






















;; run http-transport:keep-running in a parallel thread to monitor that the db is being 
;; used and to shutdown after sometime if it is not.
;;
(define (http-transport:keep-running) 
  ;; if none running or if > 20 seconds since 
  ;; server last used then start shutdown

			  (debug:print-info 0 *default-log-port* "Waiting for server alive signature")
                          (mutex-lock! *heartbeat-mutex*)
                          (set! sdat *server-info*)
                          (mutex-unlock! *heartbeat-mutex*)
                          (if (and sdat
				   (not changed)
				   (> (- (current-seconds) start-time) 2))
			      (let* ((servinfodir (server:get-servinfo-dir *toppath*)) ;; (conc *toppath*"/.servinfo"))
				     (ipaddr      (car sdat))
				     (port        (cadr sdat))
				     (servinf     (conc servinfodir"/"ipaddr":"port)))
				(set! servinfofile servinf)
				(if (not (file-exists? servinfodir))
				    (create-directory servinfodir #t))
				(with-output-to-file servinf
				  (lambda ()
				    (let* ((serv-id (server:mk-signature)))
				      (set! *server-id* serv-id)
				      (print "SERVER STARTED: "ipaddr":"port" AT "(current-seconds)" server-id: "serv-id" pid: "(current-process-id))
				      (print "started: "(seconds->year-week/day-time (current-seconds))))))
				(set! *on-exit-procs* (cons
						       (lambda ()
							 (delete-file* servinf))
						       *on-exit-procs*))
				;; put data about this server into a simple flat file host.port
				(debug:print-info 0 *default-log-port* "Received server alive signature")






				sdat)
                              (begin
				(debug:print-info 0 *default-log-port* "Still waiting, last-sdat=" last-sdat)
                                (sleep 4)
				(if (> (- (current-seconds) start-time) 120) ;; been waiting for two minutes
                                    (if sdat 
				      (let* ((ipaddr  (car sdat))
					   (port    (cadr sdat))
					   (servinf (conc (server:get-servinfo-dir *toppath*)"/"ipaddr":"port)))
				        (debug:print-error 0 *default-log-port* "transport appears to have died, exiting server")




				      (exit))



                                      (exit)
                                    )
				    (loop start-time
					  (equal? sdat last-sdat)
					  sdat)))))))
	 (iface       (car server-info))
         (port        (cadr server-info))
         (last-access 0)
	 (server-timeout (server:expiration-timeout))

	    (set! server-going #t)
	    (debug:print 0 *default-log-port* "SERVER: running, megatest version: " (common:get-full-version))) ;; NOTE: the server is NOT yet marked as running in the log. We do that in the keep-running routine.
	  (if (and no-sync-db
		   (common:low-noise-print 10 "sync-all")) ;; cheesy way to reduce frequency of running sync :)
              (begin
		(if (common:low-noise-print 120 "sync-all-print")
                    (debug:print 0 *default-log-port* "keep-running calling db:all-db-sync at " (time->string (seconds->local-time) "%H:%M:%S")))
		(db:all-db-sync *dbstruct-dbs*)
		)))
      
      ;; when things go wrong we don't want to be doing the various queries too often
      ;; so we strive to run this stuff only every four seconds or so.
      (let* ((sync-time (- (current-milliseconds) start-time))
	    (rem-time  (quotient (- 4000 sync-time) 1000)))
	(if (and (<= rem-time 4)
		 (>  rem-time 0))

		  (current-seconds)))
          (if (common:low-noise-print 120 "server continuing")
              (debug:print-info 0 *default-log-port* "Server continuing, seconds since last db access: " (- (current-seconds) last-access))
	      (let ((curr-time (current-seconds)))
		(handle-exceptions
		    exn
		    (debug:print 0 *default-log-port* "ERROR: Failed to change timestamp on info file " servinfofile ". Are you out of space on that disk? exn=" exn)
		    (if (and ;; (< (- (current-seconds) server-start-time) 600) ;; run for ten minutes for experiment, 3600 thereafter
			     (not *server-overloaded*)
			     (file-exists? servinfofile))
			(change-file-times servinfofile curr-time curr-time)))
		(if (and (common:low-noise-print 120 "start new server")
			(> *api-process-request-count* 50)) ;; if this server is kind of busy start up another
		    (begin
		      (debug:print-info 0 *default-log-port* "Server is busy, api-count "*api-process-request-count*", start another if possible...")
		      (server:kind-run *toppath*)
		      (if (> *api-process-request-count* 100)
			  (begin
			    (debug:print-info 0 *default-log-port* "Server is overloaded at api-count=" *api-process-request-count*", removing "servinfofile) 
			    (delete-file* servinfofile)))))))
          (loop 0 server-state bad-sync-count (current-milliseconds)))
         (else
          (debug:print-info 0 *default-log-port* "Server timed out. seconds since last db access: " (- (current-seconds) last-access))
          (http-transport:server-shutdown port)))))))

(define (http-transport:server-shutdown port)

    (exit)))

;; all routes though here end in exit ...
;;
;; start_server? 
;;
(define (http-transport:launch)
  ;; check the .servinfo directory, are there other servers running on this
  ;; or another host?
  (let* ((server-start-is-ok  (server:minimal-check *toppath*)))















    (if (not server-start-is-ok)
	(begin






	  (debug:print 0 *default-log-port* "ERROR: server start not ok, exiting now.")
	  (exit 1))))
    
  ;; check that a server start is in progress, pause or exit if so


  (let* ((th2 (make-thread (lambda ()
                             (debug:print-info 0 *default-log-port* "Server run thread started")
                             (http-transport:run 
                              (if (args:get-arg "-server")
                                  (args:get-arg "-server")
                                  "-")
                              )) "Server run"))
         (th3 (make-thread (lambda ()
                             (debug:print-info 0 *default-log-port* "Server monitor thread started")
                             (http-transport:keep-running)
                             "Keep running"))))
    (thread-start! th2)
    (thread-sleep! 0.25) ;; give the server time to settle before starting the keep-running monitor.
    (thread-start! th3)
    (set! *didsomething* #t)
    (thread-join! th2)
    (exit)))

;; (define (http-transport:server-signal-handler signum)
;;   (signal-mask! signum)
;;   (handle-exceptions
;;    exn
;;    (debug:print 0 *default-log-port* " ... exiting ...")
;;    (let ((th1 (make-thread (lambda ()

;;     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

;; Always use two or four digit decimal
;; 1.01, 1.02...1.10,1.11,1.1101 ... 1.99,2.00..

;; (declare (unit megatest-version))

(define megatest-version 1.8006)

		      (original-exit exit-code)))))

;; for some switches always print the command to stderr
;;
(if (args:any? "-run" "-runall" "-remove-runs" "-set-state-status" "-kill-runs" "-kill-rerun")
    (debug:print 0 *default-log-port* (string-intersperse (argv) " ")))















;;======================================================================
;; Misc setup stuff
;;======================================================================

(debug:setup)

;; Start the server - can be done in conjunction with -runall or -runtests (one day...)
;;   we start the server if not running else start the client thread
;;======================================================================

;; Server? Start up here.
;;
(if (args:get-arg "-server")
    (let ((tl        (launch:setup)))

      ;; (server:launch 0 'http)
      (http-transport:launch)
      (set! *didsomething* #t)))

;; 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")
    (begin

(if (args:get-arg "-cleanup-db")
    (begin
      (if (not (launch:setup))
	  (begin
	    (debug:print 0 *default-log-port* "Failed to setup, exiting") 
	    (exit 1)))
      (if (not (server:choose-server *toppath* 'home?))
	  (begin
	    (debug:print 0 *default-log-port* "Servers are not running on this host or no servers alive. Cannot run cleanup-db")
	    (exit 1)))
      (let ((dbstructs (db:setup #f)))
        (common:cleanup-db dbstructs))
      (set! *didsomething* #t)))

(if (args:get-arg "-mark-incompletes")
    (begin
      (if (not (launch:setup))

((license "LGPL")
 (version 0.1)
 (category misc)
 (dependencies srfi-69 srfi-1)
 (author "Matt Welland")
 (synopsis "Primitive argument processor.")
 (components (extension mtargs)))

(module mtargs
    (
     arg-hash
     get-arg
     get-arg-from

     get-args
     usage
     print-args
     any-defined?

     )

(import scheme) ;; gives us cond-expand in chicken-4

(cond-expand
 (chicken-5
  (import scheme (chicken base) (chicken port) (chicken file) (chicken process-context))
  (import srfi-69 srfi-1))
 (chicken-4
  (import chicken posix srfi-69 srfi-1))
 (else))

(define usage (make-parameter print))
(define arg-hash (make-hash-table))


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

(define (any-defined? . args)
  (not (null? (filter (lambda (x) x)
		      (map get-arg args)))))

;; (define any any-defined?)

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

	    (if (null? tail) remtargs
		(loop (car tail)(cdr tail) remtargs)))
	   (else
	    (if (null? tail)(append remtargs (list arg)) ;; return the non-used args
		(loop (car tail)(cdr tail)(append remtargs (list arg))))))))
    ))

(define (print-args arg-hash)

  (for-each (lambda (arg)
	      (print "   " arg "   " (hash-table-ref/default arg-hash arg #f)))
	    (hash-table-keys arg-hash)))


)

	 (avail  (tasks:wait-on-journal fname 10))) ;; wait up to about 10 seconds for the journal to go away
    (handle-exceptions
     exn
     (begin
       ;; (release-dot-lock fname)
       (debug:print-error 0 *default-log-port* "portlogger:open-run-close failed. " proc " " params)
       (debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
       ;; (debug:print 5 *default-log-port* "exn=" (condition->list exn))
       (if (common:file-exists? fname)(delete-file fname)) ;; brutally get rid of it
       (print-call-chain (current-error-port)))
     (let* (;; (lock   (obtain-dot-lock fname 2 9 10))
	    (db     (portlogger:open-db fname))
	    (res    (apply proc db params)))
       (sqlite3:finalize! db)
       ;; (release-dot-lock fname)

;;======================================================================
;;  S U P P O R T   F U N C T I O N S
;;======================================================================

;; if a server is either running or in the process of starting call client:setup
;; else return #f to let the calling proc know that there is no server available
;;
(define (rmt:get-connection-info areapath runremote) ;; TODO: push areapath down.

  (let* ((cinfo     (if (remote? runremote)
			(remote-conndat runremote)
			#f)))
	  (if cinfo
	      cinfo
	      (if (server:check-if-running areapath)
		  (client:setup areapath runremote)
		  #f))))

(define (rmt:on-homehost? runremote)
  (let* ((hh-dat (remote-hh-dat runremote)))
    (if (pair? hh-dat)
	(cdr hh-dat)
	(begin

     ;; also, the expire-time calculation might not be correct. We want, time-since-last-server-access > (server:get-timeout)
     ;;
     ;;DOT CASE4 [label="reset\nconnection"];
     ;;DOT MUTEXLOCK -> CASE4 [label="have connection,\nlast_access > expire_time"]; {rank=same "case 4" CASE4}
     ;;DOT CASE4 -> "rmt:send-receive";
     ;; reset the connection if it has been unused too long
     ((and runremote
           ;; (remote-conndat runremote)
	   (> (current-seconds) ;; if it has been more than server-timeout seconds since last contact, close this connection and start a new on
	      (+ (remote-last-access runremote)
		 (remote-server-timeout runremote))))
      (debug:print-info 0 *default-log-port* "Connection to " (remote-server-url runremote) " expired due to no accesses, forcing new connection.")
      (http-transport:close-connections runremote)
      ;; moving this setting of runremote conndat to #f to inside the http-transport:close-connections
      ;; (remote-conndat-set! runremote #f) ;; invalidate the connection, thus forcing a new connection.

      (mutex-unlock! *rmt-mutex*)
      (rmt:send-receive cmd rid params attemptnum: attemptnum))
     
     ;;DOT CASE5 [label="local\nread"];
     ;;DOT MUTEXLOCK -> CASE5 [label="server not required,\non homehost,\nread-only query"]; {rank=same "case 5" CASE5};
     ;;DOT CASE5 -> "rmt:open-qry-close-locally";

     ;;DOT CASE6 [label="init\nremote"];
     ;;DOT MUTEXLOCK -> CASE6 [label="on homehost,\nwrite query,\nhave server,\ncan't reach it"]; {rank=same "case 6" CASE6};
     ;;DOT CASE6 -> "rmt:send-receive";
     ;; on homehost and this is a write, we already have a server, but server has died

     ;; reinstate this keep-alive section but inject a time condition into the (add ...
     ;;
     ;; ((and (cdr (remote-hh-dat runremote))           ;; on homehost
     ;;       (not (member cmd api:read-only-queries))  ;; this is a write
     ;;       (remote-server-url runremote)             ;; have a server
     ;;       (not (server:ping (remote-server-url runremote) (remote-server-id runremote))))  ;; server has died. NOTE: this is not a cheap call! Need better approach.
     ;;  (debug:print 0 *default-log-port* "WARNING: server appears to have died, trying to reconnect, case 6")
     ;;  (http-transport:close-connections area-dat: runremote) ;; make sure to clean up
     ;;  (set! *runremote* (make-remote))
     ;;  (let* ((server-info (remote-server-info *runremote*))) 
     ;;        (if server-info
     ;; 		(begin
     ;; 		  (remote-server-url-set! *runremote* (server:record->url server-info))
     ;;              (remote-server-id-set! *runremote* (server:record->id server-info)))))
     ;;  (remote-force-server-set! runremote (common:force-server?))
     ;;  (mutex-unlock! *rmt-mutex*)
     ;;  (debug:print-info 12 *default-log-port* "rmt:send-receive, case  6")
     ;;  (rmt:send-receive cmd rid params attemptnum: attemptnum))

     ;;DOT CASE7 [label="homehost\nwrite"];
     ;;DOT MUTEXLOCK -> CASE7 [label="server not required,\non homehost,\na write,\nhave a server"]; {rank=same "case 7" CASE7};
     ;;DOT CASE7 -> "rmt:open-qry-close-locally";
     ;; on homehost and this is a write, we already have a server
     ((and (not (remote-force-server runremote))     ;; honor forced use of server, i.e. server NOT required
	   (cdr (remote-hh-dat runremote))           ;; on homehost

	       (not (remote-conndat runremote)))
	  (and (not (cdr (remote-hh-dat runremote)))        ;; not on a homehost 
	       (not (remote-conndat runremote))))           ;; and no connection
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case 9, hh-dat: " (remote-hh-dat runremote) " conndat: " (remote-conndat runremote))
      (mutex-unlock! *rmt-mutex*)
      (if (not (server:check-if-running *toppath*)) ;; who knows, maybe one has started up?
	  (server:start-and-wait *toppath*))
      (remote-conndat-set! runremote (rmt:get-connection-info *toppath* runremote)) ;; calls client:setup which calls client:setup-http
      (rmt:send-receive cmd rid params attemptnum: attemptnum)) ;; TODO: add back-off timeout as

     ;;DOT CASE10 [label="on homehost"];
     ;;DOT MUTEXLOCK -> CASE10 [label="server not required,\non homehost"]; {rank=same "case 10" CASE10};
     ;;DOT CASE10 -> "rmt:open-qry-close-locally";
     ;; all set up if get this far, dispatch the query
     ((and (not (remote-force-server runremote))
	   (cdr (remote-hh-dat runremote))) ;; we are on homehost
      (mutex-unlock! *rmt-mutex*)
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case 10")
      (rmt:open-qry-close-locally cmd (if rid rid 0) params))

     ;;DOT CASE11 [label="send_receive"];
     ;;DOT MUTEXLOCK -> CASE11 [label="else"]; {rank=same "case 11" CASE11};
     ;;DOT CASE11 -> "rmt:send-receive" [label="call failed"];
     ;;DOT CASE11 -> "RESULT" [label="call succeeded"];
     ;; not on homehost, do server query
     (else (extras-case-11 *default-log-port* runremote cmd params attemptnum rid)))))
    ;;DOT }























;; bunch of small functions factored out of send-receive to make debug easier
;;

(define (extras-case-11 *default-log-port* runremote cmd params attemptnum rid)
  ;; (mutex-unlock! *rmt-mutex*)
  (debug:print-info 12 *default-log-port* "rmt:send-receive, case  9")
  ;; (mutex-lock! *rmt-mutex*)
  (let* ((conninfo (remote-conndat runremote))

	 (dat-in  (condition-case ;; handling here has
			     ;; caused a lot of
			     ;; problems. However it
			     ;; is needed to deal with
			     ;; attemtped
			     ;; communication to
			     ;; servers that have gone
			     ;; away
			     (http-transport:client-api-send-receive 0 runremote cmd params)
			     ;; (http-transport:client-api-send-receive 0 conninfo cmd params runremote)
			     ((servermismatch)  (vector #f "Server id mismatch" ))
			     ((commfail)(vector #f "communications fail"))
			     ((exn)(vector #f "other fail" (print-call-chain)))))



	 (dat      (if (and (vector? dat-in) ;; ... check it is a correct size
			    (> (vector-length dat-in) 1))
		       dat-in
		       (vector #f (conc "communications fail (type 2), dat-in=" dat-in))))
	 (success  (if (vector? dat) (vector-ref dat 0) #f))
	 (res      (if (vector? dat) (vector-ref dat 1) #f)))
    (if (and (vector? conninfo) (< 5 (vector-length conninfo)))
	(remote-last-access-set! runremote (current-seconds)) ;; refresh access time
	(begin
	  (debug:print 0 *default-log-port* "INFO: Should not get here! conninfo=" conninfo)
	  (set! conninfo #f)

	  (http-transport:close-connections runremote)))
    (debug:print-info 13 *default-log-port* "rmt:send-receive, case  9. conninfo=" conninfo " dat=" dat " runremote = " runremote)
    (mutex-unlock! *rmt-mutex*)
    (if success ;; success only tells us that the transport was
	;; successful, have to examine the data to see if
	;; there was a detected issue at the other end
	(extras-transport-succeded *default-log-port* *rmt-mutex* attemptnum runremote res params rid cmd)
	(begin

	  (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 runremote cmd run-id params)
  (let* ((run-id   (if run-id run-id 0))





	 (res  	   (http-transport:client-api-send-receive run-id runremote cmd params)))
    (if (and res (vector-ref res 0))
	(vector-ref res 1) ;;; YES!! THIS IS CORRECT!! CHANGE IT HERE, THEN CHANGE rmt:send-receive ALSO!!!
	#f)))

;;======================================================================
;;
;; A C T U A L   A P I   C A L L S  

(define (rmt:login run-id)
  (rmt:send-receive 'login run-id (list *toppath* megatest-version (client:get-signature))))

;; This login does no retries under the hood - it acts a bit like a ping.
;; Deprecated for nmsg-transport.
;;
(define (rmt:login-no-auto-client-setup runremote)

  (rmt:send-receive-no-auto-client-setup runremote 'login 0 (list *toppath* megatest-version (client:get-signature))))



;; hand off a call to one of the db:queries statements
;; added run-id to make looking up the correct db possible 
;;
(define (rmt:general-call stmtname run-id . params)
  (rmt:send-receive 'general-call run-id (append (list stmtname run-id) params)))

  (debug:print-info 12 log-port "rmt:send-receive, case 3")
  (debug:print 0 log-port "WARNING: write transaction requested on a readonly area.  cmd="cmd" params="params)
  #f)

(define (extras-transport-failed *default-log-port* *rmt-mutex* attemptnum runremote cmd rid params)
  (debug:print 0 *default-log-port* "WARNING: communication failed. Trying again, try num: " attemptnum)
  (mutex-lock! *rmt-mutex*)

  (http-transport:close-connections runremote)
  (remote-server-url-set! runremote #f)
  (mutex-unlock! *rmt-mutex*)
  (debug:print-info 12 *default-log-port* "rmt:send-receive, case  9.1")
  (rmt:send-receive cmd rid params attemptnum: (+ attemptnum 1)))
  
(define (extras-transport-succeded *default-log-port* *rmt-mutex* attemptnum runremote res params rid cmd)
  (if (and (vector? res)

						 ;; server is
						 ;; overloaded and we
						 ;; want to ease off
						 ;; the queries
      (let ((wait-delay (+ attemptnum (* attemptnum 10))))
	(debug:print 0 *default-log-port* "WARNING: server is overloaded. Delaying " wait-delay " seconds and trying call again.")
	(mutex-lock! *rmt-mutex*)
	(http-transport:close-connections runremote)
	(set! *runremote* #f) ;; force starting over
	(mutex-unlock! *rmt-mutex*)
	(thread-sleep! wait-delay)
	(rmt:send-receive cmd rid params attemptnum: (+ attemptnum 1)))
      res)) ;; All good, return res

#;(set-functions rmt:send-receive                       remote-server-url-set!
	       http-transport:close-connections	      remote-conndat-set!
	       debug:print                            debug:print-info
	       remote-ro-mode                         remote-ro-mode-set!
	       remote-ro-mode-checked-set!            remote-ro-mode-checked)

			 ", "))
      (thread-sleep! 0.051)
      (list hed tal reg reruns))
     
     ;; 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" 600)
	  (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! 0.25)
      
      ;; new logic.
      ;; If it has been more than 10 seconds since we were last here don't wait at all
      ;; otherwise sleep 2 seconds to give db a rest and let dashboard read data
      (if (runs:lownoise "frequent-no-resources" 10)
	  (thread-sleep! 0.25) ;; no significant delay
	  (thread-sleep! 2))
      ;; 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)

		 (let* ((jobgroup              (runs:testdat-jobgroup testdat-in))
			(can-run-more-tests    (runs:dat-can-run-more-tests runsdat))
			(last-jobs-check-time  (runs:dat-last-jobs-check-time runsdat))
			(should-check-jobs     (match can-run-more-tests
						 ((can-run-more-flag num-running nr-in-jobgroup max-concurrent-jobs . params)
						  (if (< (- max-concurrent-jobs num-running) 25)
						      (begin
							(debug:print-info 2 *default-log-port*
									  "less than 20 jobs headroom, ("max-concurrent-jobs
									  "-"num-running")>20. Forcing prelaunch check.")
							#t)
						      #f))
						 (else #f)))) ;; no record yet
		   (if should-check-jobs
		       (let loop-can-run-more

;;======================================================================
;; S E R V E R
;;======================================================================

;; Call this to start the actual server
;;












;;======================================================================
;; S E R V E R   U T I L I T I E S 
;;======================================================================

;; Get the transport
(define (server:get-transport)
  (if *transport-type*

;; Given an area path,  start a server process    ### NOTE ### > file 2>&1 
;; if the target-host is set 
;; try running on that host
;;   incidental: rotate logs in logs/ dir.
;;
(define  (server:run areapath) ;; areapath is *toppath* for a given testsuite area







  (let* ((testsuite   (common:get-testsuite-name))
	 (logfile     (conc areapath "/logs/server.log")) ;; -" curr-pid "-" target-host ".log"))
	 (profile-mode (or (configf:lookup *configdat* "misc" "profilesw")
			   ""))
	 (cmdln (conc (common:get-megatest-exe)
		      " -server - ";; (or target-host "-")
		      (if (equal? (configf:lookup *configdat* "server" "daemonize") "yes")
			  " -daemonize "

(define (server:logf-get-start-info logf)
  (let ((server-rx    (regexp "^SERVER STARTED: (\\S+):(\\d+) AT ([\\d\\.]+) server-id: (\\S+) pid: (\\d+)")) ;; SERVER STARTED: host:port AT timesecs server id
        (dbprep-rx    (regexp "^SERVER: dbprep"))
        (dbprep-found 0)
	(bad-dat      (list #f #f #f #f #f)))
    (handle-exceptions
     exn
     (begin
       ;; WARNING: this is potentially dangerous to blanket ignore the errors
       (if (file-exists? logf)
	   (debug:print-info 2 *default-log-port* "Unable to get server info from "logf", exn=" exn))
       bad-dat) ;; no idea what went wrong, call it a bad server
     (with-input-from-file
	 logf
       (lambda ()
	 (let loop ((inl  (read-line))
		    (lnum 0))
	   (if (not (eof-object? inl))
	       (let ((mlst (string-match server-rx inl))
		     (dbprep (string-match dbprep-rx inl)))
		 (if dbprep (set! dbprep-found 1))
		 (if (not mlst)
		     (if (< lnum 500) ;; give up if more than 500 lines of server log read
			 (loop (read-line)(+ lnum 1))
			 (begin 
                           (debug:print-info 0 *default-log-port* "Unable to get server info from first 500 lines of " logf )
                           bad-dat))
		     (match mlst
			    ((_ host port start server-id pid)
			     (list host
				   (string->number port)
				   (string->number start)
				   server-id
				   (string->number pid)))
			    (else
			     (debug:print 0 *current-log-port* "ERROR: did not recognise SERVER line info "mlst)
			     bad-dat))))
	       (begin 
		 (if dbprep-found
		     (begin
		       (debug:print-info 2 *default-log-port* "Server is in dbprep at " (common:human-time))
		       (thread-sleep! 0.5)) ;; was 25 sec but that blocked things from starting?
		     (debug:print-info 0 *default-log-port* "Unable to get server info from " logf " at " (seconds->time-string (current-seconds))))
		 bad-dat))))))))

;; ;; get a list of servers from the log files, with all relevant data
;; ;; ( mod-time host port start-time pid )
;; ;;
;; (define (server:get-list areapath #!key (limit #f))
;;   (let ((fname-rx    (regexp "^(|.*/)server-(\\d+)-(\\S+).log$"))
;; 	(day-seconds (* 24 60 60)))

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

;; oldest server alive determines host then choose random of youngest
;; five servers on that host
;;
(define (server:get-servers-info areapath)
  ;; (assert *toppath* "FATAL: server:get-servers-info called before *toppath* has been set.")
  (let* ((servinfodir (server:get-servinfo-dir areapath))) ;; (conc *toppath*"/.servinfo")))
    (if (not (file-exists? servinfodir))
	(create-directory servinfodir))
    (let* ((allfiles    (glob (conc servinfodir"/*")))
	   (res         (make-hash-table)))
      (for-each
       (lambda (f)
	 (let* ((hostport  (pathname-strip-directory f))
		(serverdat (server:logf-get-start-info f)))
	   (match serverdat
	     ((host port start server-id pid)
	      (if (and host port start server-id pid)
		  (hash-table-set! res hostport serverdat)
		  (debug:print-info 2 *default-log-port* "bad server info for "f": "serverdat)))
	     (else
	      (debug:print-info 2 *default-log-port* "bad server info for "f": "serverdat)))))
       allfiles)
      res)))

;; check the .servinfo directory, are there other servers running on this
;; or another host?
;;
;; returns #t => ok to start another server
;;         #f => not ok to start another server
;;
(define (server:minimal-check areapath)
  (server:clean-up-old areapath)
  (let* ((srvdir      (server:get-servinfo-dir areapath)) ;; (conc areapath"/.servinfo"))
	 (servrs      (glob (conc srvdir"/*")))
	 (thishostip  (server:get-best-guess-address (get-host-name)))
	 (thisservrs  (glob (conc srvdir"/"thishostip":*")))
	 (homehostinf (server:choose-server areapath 'homehost))
	 (havehome    (car homehostinf))
	 (wearehome   (cdr homehostinf)))
    (debug:print-info 0 *default-log-port* thishostip", have homehost: "havehome", we are homehost: "wearehome
		      ", numservers: "(length thisservrs))
    (cond
     ((not havehome) #t) ;; no homehost yet, go for it
     ((and havehome wearehome (< (length thisservrs) 20)) #t) ;; we are home and less than 20 servers, ok to start another
     ((and havehome (not wearehome)) #f)     ;; we are not the home host
     ((and havehome wearehome (>= (length thisservrs) 20)) #f) ;; have enough running
     (else
      (debug:print 0 *default-log-port* "WARNING: Unrecognised scenario, servrs="servrs", thishostip="thishostip", thisservrs="thisservrs)
      #t))))
	 

(define server-last-start 0)


;; oldest server alive determines host then choose random of youngest
;; five servers on that host
;;
;; mode:
;;   best - get best server (random of newest five)
;;   home - get home host based on oldest server
;;   info - print info
(define (server:choose-server areapath #!optional (mode 'best))
  ;; age is current-starttime
  ;; find oldest alive
  ;;   1. sort by age ascending and ping until good
  ;; find alive rand from youngest
  ;;   1. sort by age descending
  ;;   2. take five
  ;;   3. check alive, discard if not and repeat
  ;; first we clean up old server files
  (server:clean-up-old areapath)
  (let* ((since-last (- (current-seconds) server-last-start))
        (server-start-delay 10))     
    (if ( < (- (current-seconds) server-last-start) 10 )
      (begin
        (debug:print 2 *default-log-port* "server:choose-server: seconds since last server start: " (- (current-seconds) server-last-start))
        (debug:print 2 *default-log-port* "server:choose-server: last server start less than " server-start-delay " seconds ago. Sleeping " server-start-delay " seconds")
        (thread-sleep! server-start-delay)
      )
      (debug:print 2 *default-log-port* "server:choose-server: seconds since last server start: " (- (current-seconds) server-last-start))
    )
  )
  (let* ((serversdat  (server:get-servers-info areapath))
	 (servkeys    (hash-table-keys serversdat))
	 (by-time-asc (if (not (null? servkeys)) ;; NOTE: Oldest is last
			  (sort servkeys ;; list of "host:port"
				(lambda (a b)
				  (>= (list-ref (hash-table-ref serversdat a) 2)
				      (list-ref (hash-table-ref serversdat b) 2))))
			  '())))
    (debug:print 2 *default-log-port* "server:choose-server: serversdat: " serversdat)
    (debug:print 2 *default-log-port* "server:choose-server: servkeys: " servkeys)
    (if (not (null? by-time-asc))
	(let* ((oldest     (last by-time-asc))
	       (oldest-dat (hash-table-ref serversdat oldest))
	       (host       (list-ref oldest-dat 0))
	       (all-valid  (filter (lambda (x)
				     (equal? host (list-ref (hash-table-ref serversdat x) 0)))
				   by-time-asc))
	       (best-ten  (lambda ()
			     (if (> (length all-valid) 11)
				 (take (drop-right all-valid 1) 10) ;; remove the oldest from consideration so it can age out
				 (if (> (length all-valid) 8)
				     (drop-right all-valid 1)
				     all-valid))))
	       (names->dats (lambda (names)
			      (map (lambda (x)
				     (hash-table-ref serversdat x))
				   names)))
	       (am-home?    (lambda ()
			      (let* ((currhost (get-host-name))
				     (bestadrs (server:get-best-guess-address currhost)))
				(or (equal? host currhost)
				    (equal? host bestadrs))))))
	  (case mode
	    ((info)
	     (print "oldest: "oldest-dat", selected host: "host", all-valid: "all-valid)
	     (print "youngest: "(hash-table-ref serversdat (car all-valid))))
	    ((home)     host)
	    ((homehost) (cons host (am-home?))) ;; shut up old code
	    ((home?)    (am-home?))
	    ((best-ten)(names->dats (best-ten)))
	    ((all-valid)(names->dats all-valid))
	    ((best)     (let* ((best-ten (best-ten))
			       (len       (length best-ten)))
			  (hash-table-ref serversdat (list-ref best-ten (random len)))))
	    ((count)(length all-valid))
	    (else
	     (debug:print 0 *default-log-port* "ERROR: invalid command "mode)
	     #f)))
	(begin
	  (server:run areapath)
          (set! server-last-start (current-seconds))
	  ;; (thread-sleep! 3)
	  (case mode
	    ((homehost) (cons #f #f))
	    (else	#f))))))

(define (server:get-servinfo-dir areapath)
  (let* ((spath (conc areapath"/.servinfo")))
    (if (not (file-exists? spath))
	(create-directory spath #t))
    spath))

(define (server:clean-up-old areapath)
  ;; any server file that has not been touched in ten minutes is effectively dead
  (let* ((sfiles (glob (conc (server:get-servinfo-dir areapath)"/*"))))
    (for-each
     (lambda (sfile)
       (let* ((modtime (handle-exceptions
			   exn
			 (begin
			   (debug:print 0 *default-log-port* "WARNING: failed to get modification file for "sfile)
			   (current-seconds))
			 (file-modification-time sfile))))
	 (if (and (number? modtime)
		  (> (- (current-seconds) modtime)
		     600))
	     (begin
	       (debug:print 0 *default-log-port* "WARNING: found old server info file "sfile", removing it.")
	       (handle-exceptions
		   exn
		 (debug:print 0 *default-log-port* "WARNING: failed to delete old server info file "sfile)
		 (delete-file sfile))))))
     sfiles)))

;; would like to eventually get rid of this
;;
(define (common:on-homehost?)
  (server:choose-server *toppath* 'home?))

;; 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)
  (let loop ()
    (if (> (alist-ref 'adj-proc-load (common:get-normalized-cpu-load #f)) 2)
	(begin
	  (if (common:low-noise-print 30 "our-host-load")
	      (debug:print 0 *default-log-port* "WARNING: system load is high, waiting to start server."))
	  (loop))))
  (if (< (server:choose-server areapath 'count) 20)
      (server:run 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  2 *default-log-port* "server:kind-run: touching " start-flag)
	  (system (conc "touch " start-flag)) ;; lazy but safe

(define (server:start-and-wait areapath #!key (timeout 60))
  (let ((give-up-time (+ (current-seconds) timeout)))
    (let loop ((server-info (server:check-if-running areapath))
	       (try-num    0))
      (if (or server-info
	      (> (current-seconds) give-up-time)) ;; server-url will be #f if no server available.
	  (server:record->url server-info)
	  (let* ( (servers (server:choose-server areapath 'all-valid))
                (num-ok (if servers (length (server:choose-server areapath 'all-valid)) 0)))
	    (if (and (> try-num 0)  ;; first time through simply wait a little while then try again
		     (< num-ok 1))  ;; if there are no decent candidates for servers then try starting a new one
		(server:run areapath))
	    (thread-sleep! 5)
	    (loop (server:check-if-running areapath)
		  (+ try-num 1)))))))

(define (server:get-num-servers #!key (numservers 2))
  (let ((ns (string->number
	     (or (configf:lookup *configdat* "server" "numservers") "notanumber"))))
    (or ns numservers)))

;; no longer care if multiple servers are started by accident. older servers will drop off in time.
;;
(define (server:check-if-running areapath) ;;  #!key (numservers "2"))
  (let* ((ns            (server:get-num-servers)) ;; get the setting the for maximum number of servers allowed
	 (servers       (server:choose-server areapath 'best-ten))) ;; (server:get-best (server:get-list areapath))))
    (if (or (and servers
		 (null? servers))
	    (not servers))
	    ;; (and (list? servers)
	    ;;	 (< (length servers) (+ 1 (random ns))))) ;; somewhere between 1 and numservers
        #f
        (let loop ((hed (car servers))

    (tasks:kill-server hostname pid))))

;; called in megatest.scm, host-port is string hostname:port
;;
;; NOTE: This is NOT called directly from clients as not all transports support a client running
;;       in the same process as the server.
;;
(define (server:ping host:port server-id #!key (do-exit #f))









  (let* ((host-port (cond
		     ((string? host:port)
		      (let ((slst (string-split   host:port ":")))
			(if (eq? (length slst) 2)
			    (list (car slst)(string->number (cadr slst)))
			    #f)))
		     (else
		      #f))))


    (cond
     ((and (list? host-port)

	   (eq? (length host-port) 2))



      (let* ((myrunremote (make-remote))
	     (iface       (car host-port))
	     (port        (cadr host-port))
	     (server-dat  (client:connect iface port server-id myrunremote))
	     (login-res   (rmt:login-no-auto-client-setup myrunremote)))
	(if (and (list? login-res)
		 (car login-res))
	    (begin
	      ;; (print "LOGIN_OK")
	      (if do-exit (exit 0))
	      #t)
	    (begin
	      ;; (print "LOGIN_FAILED")
	      (if do-exit (exit 1))
	      #f))))
     (else 
      (if host:port
	  (debug:print 0 *default-log-port*  "ERROR: bad host:port "host:port))
      (if do-exit
	  (exit 1)
	  #f)))))

;; run ping in separate process, safest way in some cases
;;
(define (server:ping-server ifaceport)
  (with-input-from-pipe 
   (conc (common:get-megatest-exe) " -ping " ifaceport)
   (lambda ()

;; Default is 60 seconds.
;;
(define (server:expiration-timeout)
  (let ((tmo (configf:lookup *configdat* "server" "timeout")))
    (if (and (string? tmo)
	     (common:hms-string->seconds tmo)) ;; BUG: hms-string->seconds is broken, if given "10" returns 0. Also, it doesn't belong in this logic unless the string->number is changed below
        (* 3600 (string->number tmo))
	600)))

(define (server:get-best-guess-address hostname)
  (let ((res #f))
    (for-each 
     (lambda (adr)
       (if (not (eq? (u8vector-ref adr 0) 127))
	   (set! res adr)))