Megatest

(include "db_records.scm")

;; these come from processmod
;;
;; (define setenv set-environment-variable!)
;; (define unsetenv unset-environment-variable!)
;; (define getenv get-environment-variable)

;; move all the miscellanea into this struct
;;
(defstruct runs:gendat inc-results inc-results-last-update inc-results-fmt run-info runname target)


(define *bdat* #f) ;; the one and only (someday) global?

(defstruct bdat
  (home                   (getenv "HOME"))
  (user                   (getenv "USER"))
  (watchdog               #f)
  (time-to-exit           #f)
  (task-db                #f)
  (target                 #f)
  (this-exe-fullpath      #f)
  (this-exe-dir           #f)
  (this-exe-name          #f)
  (orig-env               #f)

  ;; runs stuff
  (runs-data              #f) ;; was runs:general-data
  ;; (server-loop-heart-beat (current-seconds))
  )

;; move all needed initialization into here
;; break it into pieces if need be later
;;
(define (make-and-init-bigdata)
  (let* ((bdat 	(make-bdat))
	 (fullp (common:get-this-exe-fullpath)))
    ;; bdat stuff
    (bdat-this-exe-fullpath-set! bdat fullp)
    (bdat-this-exe-dir-set! bdat (pathname-directory fullp))
    (bdat-this-exe-name-set! bdat (pathname-strip-directory fullp))
    (bdat-orig-env-set! bdat (get-the-original-environment))
    ;; setup runs-data
    (bdat-runs-data-set! bdat  (make-runs:gendat
				inc-results: (make-hash-table)
				inc-results-last-update: 0
				;;               state status time duration test-name item-path
				inc-results-fmt: "~12a~12a~20a~12a~40a\n"
				run-info: #f
				runname: #f
				target: #f))
    
    (set! *bdat* bdat)
    ;; set up signal handlers
    (set-signal-handler! signal/int  std-signal-handler)  ;; ^C
    (set-signal-handler! signal/term std-signal-handler)
    ;; (set-signal-handler! signal/stop std-signal-handler)  ;; ^Z NO, do NOT handle ^Z!
    bdat))

;; D E B U G G I N G   S T U F F 
;;======================================================================

(define *verbosity*         1)
(define *logging*           #f)

(define *common:thread-punchlist* (make-hash-table))

(define *last-num-running-tests* 0)
(define *seen-cant-run-tests* (make-hash-table)) ;; use to track tests that we suspect cannot be run
(define *runs:denoise* (make-hash-table)) ;; key => last-time-ran

(define runs:nothing-left-in-queue-count 0)

(define *max-tries-hash* (make-hash-table))

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

(define *db:process-queue-mutex* (make-mutex))
(define *http-functions* (make-hash-table))
(define *http-mutex* (make-mutex))

;; NOTE: Large block of code from 32436b426188080f72fceb6894af541fbad9921e removed here
;;       I'm pretty sure it is defunct.

;; This next block all imported en-mass from the api branch
(define *http-requests-in-progress* 0)
(define *http-connections-next-cleanup* (current-seconds))

(define *number-of-writes* 0)
(define *number-non-write-queries* 0)

;; NB// #f => return dbdir only
;;      (was planned to be;  zeroth db with name=main.db)
;; 
;; If run-id is #f return to create and retrieve the path where the db will live.
;;
;; (define db:dbfile-path common:get-db-tmp-area)
(define *global-db-store* (make-hash-table))


;;======================================================================
;; end globals
;;======================================================================

;;                       0           1              2              3
(defstruct launch:einf (pid #t)(exit-status #t)(exit-code #t)(rollup-status 0))

;; 
;; (include "common_records.scm")

;; (include "db_records.scm")
(include "key_records.scm")
;; (include "run_records.scm")

(define *number-of-writes* 0)
(define *number-non-write-queries* 0)

;;======================================================================
;;  R E C O R D S
;;======================================================================

;; each db entry is a pair ( db . dbfilepath )
;; I propose this record evolves into the area record
;;

;;    if db already open - return inmem
;;    if db not open, open inmem, rundb and sync then return inmem
;;    inuse gets set automatically for rundb's
;;
(define (db:get-db dbstruct) ;;  run-id) 
  (if (stack? (dbr:dbstruct-dbstack dbstruct))
      (if (stack-empty? (dbr:dbstruct-dbstack dbstruct))
          (let ((newdb (db:open-megatest-db path: (db:dbfile-path))))
          (let ((newdb (db:open-megatest-db path: (common:get-db-tmp-area))))
            ;; (stack-push! (dbr:dbstruct-dbstack dbstruct) newdb)
            newdb)
          (stack-pop! (dbr:dbstruct-dbstack dbstruct)))
      (db:open-db dbstruct)))

;; ;; legacy handling of structure for managing db's. Refactor this into dbr:?
(define (db:dbdat-get-db dbdat)
  (if (pair? dbdat)
      (car dbdat)
      dbdat))

(define (db:dbdat-get-path dbdat)
  (if (pair? dbdat)
      (cdr dbdat)
      #f))

(define-inline (db:generic-error-printout exn . message)
(define (db:generic-error-printout exn . message)
  (print-call-chain (current-error-port))
  (apply debug:print-error 0 *default-log-port* message)
  (debug:print-error 0 *default-log-port* "   params: " params
  (debug:print-error 0 *default-log-port* ;; "   params: " params
		     ", error: "     ((condition-property-accessor 'exn 'message)   exn)
		     ", arguments: " ((condition-property-accessor 'exn 'arguments) exn)
		     ", location: "  ((condition-property-accessor 'exn 'location)  exn)
		     ))

;; (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

;; 
;; ;; Use to get a path. To get an arbitrary string see next define
;; ;;
;; (define (db:get-path dbstruct id)
;;   (let ((fdb (db:get-filedb dbstruct)))
;;     (filedb:get-path db id)))

;; NB// #f => return dbdir only
;;      (was planned to be;  zeroth db with name=main.db)
;; 
;; If run-id is #f return to create and retrieve the path where the db will live.
;;
(define db:dbfile-path common:get-db-tmp-area)

(define (db:set-sync db)
  (let ((syncprag (configf:lookup *configdat* "setup" "sychronous")))
    (sqlite3:execute db (conc "PRAGMA synchronous = " (or syncprag 0) ";")))) 

;; open an sql database inside a file lock
;; returns: db existed-prior-to-opening
;; RA => Returns a db handler; sets the lock if opened in writable mode

;; This routine creates the db if not already present. It is only called if the db is not already opened
;;
(define (db:open-db dbstruct #!key (areapath #f)(do-sync #t)) ;; TODO: actually use areapath
  (let ((tmpdb-stack (dbr:dbstruct-dbstack dbstruct))) ;; RA => Returns the first reference in dbstruct
    (if (stack? tmpdb-stack)
	(db:get-db tmpdb-stack) ;; get previously opened db (will create new db handle if all in the stack are already used
        (let* ((max-stale-tmp (configf:lookup-number *configdat* "server" "filling-db-max-stale-seconds" default: 10))
               (dbpath       (db:dbfile-path ))      ;; path to tmp db area
               (dbpath       (common:get-db-tmp-area ))      ;; path to tmp db area
               (dbexists     (common:file-exists? dbpath))
	       (tmpdbfname   (conc dbpath "/megatest.db"))
	       (dbfexists    (common:file-exists? tmpdbfname))  ;; (conc dbpath "/megatest.db")))
               (mtdbexists   (common:file-exists? (conc *toppath* "/megatest.db")))
							 
               (mtdbmodtime  (if mtdbexists (common:lazy-sqlite-db-modification-time (conc *toppath* "/megatest.db"))   #f))
	        		 (tmpdbmodtime (if dbfexists  (common:lazy-sqlite-db-modification-time tmpdbfname) #f)) 

	 ;; fill in blanks (not allowed as it would be part of the path
	 (sqlite3:execute
	  db
	  (conc "UPDATE runs SET " column "='x' WHERE " column "='';"))
	 (set! field-num (+ field-num 1))))
     fields)))
  
(define *global-db-store* (make-hash-table))

(define (db:get-access-mode)
  (if (args:get-arg "-use-db-cache") 'cached 'rmt))

;; Add db direct
;;
(define (db:dispatch-query access-mode rmt-cmd db-cmd . params)
  (if (eq? access-mode 'cached)

		(sqlite3:execute db "DELETE FROM metadat WHERE var=?;" var))))

;;======================================================================
;; no-sync.db - small bits of data to be shared between servers
;;======================================================================

(define (db:open-no-sync-db)
  (let* ((dbpath (db:dbfile-path))
  (let* ((dbpath (common:get-db-tmp-area))
	 (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 "PRAGMA synchronous = 0;")

		   )))
    (debug:print-info 11 *default-log-port* "db:get-runs END qrystr: " qrystr " target: " target " offset: " offset " limit: " count)
    res))

;; TODO: Switch this to use max(update_time) from each run db? Then if using a server there is no disk traffic (using inmem db)
;;
(define (db:get-changed-run-ids since-time)
  (let* ((dbdir      (db:dbfile-path)) ;; (configf:lookup *configdat* "setup" "dbdir"))
  (let* ((dbdir      (common:get-db-tmp-area)) ;; (configf:lookup *configdat* "setup" "dbdir"))
	 (alldbs     (glob (conc dbdir "/[0-9]*.db")))
	 (changed    (filter (lambda (dbfile)
			       (> (file-modification-time dbfile) since-time))
			     alldbs)))
    (delete-duplicates
     (map (lambda (dbfile)
	    (let* ((res (string-match ".*\\/(\\d)*\\.db" dbfile)))

		 (indx 0))
	(if (equal? fieldname hed)
	    indx
	    (if (null? tal)
		#f
		(loop (car tal)(cdr tal)(+ indx 1)))))))

;; CONVERT THIS TO A FUNCTION!
(define db:test-record-qry-selector (string-intersperse db:test-record-fields ","))

(define (db:update-tesdata-on-repilcate-db dbstruct old-lt new-lt)
  (db:with-db
   dbstruct   #f   #f
   (lambda (db)
     (sqlite3:execute db "UPDATE tests SET rundir= replace(rundir,?,?), shortdir=replace(shortdir,?,?);"

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

;; Call this to start the actual server
;;

(define *db:process-queue-mutex* (make-mutex))
(define *http-functions* (make-hash-table))
(define (http-get-function fnkey)
  (hash-table-ref/default *http-functions* fnkey (lambda () "nothing here yet")))

(define (http-transport:run hostn)
  ;; Configurations for server
  (tcp-buffer-size 2048)
  (max-connections 2048) 

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

;;======================================================================
;; C L I E N T S
;;======================================================================

(define *http-mutex* (make-mutex))

;; NOTE: Large block of code from 32436b426188080f72fceb6894af541fbad9921e removed here
;;       I'm pretty sure it is defunct.

;; This next block all imported en-mass from the api branch
(define *http-requests-in-progress* 0)
(define *http-connections-next-cleanup* (current-seconds))

(define (http-transport:get-time-to-cleanup)
  (let ((res #f))
    (mutex-lock! *http-mutex*)
    (set! res (> (current-seconds) *http-connections-next-cleanup*))
    (mutex-unlock! *http-mutex*)
    res))

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

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

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

;; 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)))

				 (cons 'none 0))
			     (loop (car tal)(cdr tal) newmax-cmd currmax)))))))
    (mutex-unlock! *db-stats-mutex*)
    res))

(define (rmt:open-qry-close-locally cmd run-id params #!key (remretries 5))
  (let* ((qry-is-write   (not (member cmd api:read-only-queries)))
	 (db-file-path   (db:dbfile-path)) ;;  0))
	 (db-file-path   (common:get-db-tmp-area)) ;;  0))
	 (dbstruct-local (db:setup #t))  ;; make-dbr:dbstruct path:  dbdir local: #t)))
	 (read-only      (not (file-writable? db-file-path)))
	 (start          (current-milliseconds))
	 (resdat         (if (not (and read-only qry-is-write))
			     (let ((v (api:execute-requests dbstruct-local (vector (symbol->string cmd) params))))
			       (handle-exceptions ;; there has been a long history of receiving strange errors from values returned by the client when things go wrong..
				exn               ;;  This is an attempt to detect that situation and recover gracefully

      (debug:print-info 0 *default-log-port* "DONE: rtime=" rtime ", elapsed time=" (- done-time startt)
			", ratio=" (/ rtime (- done-time startt))))))

;; Every time can-run-more-tests is called increment the delay
;;
;; NOTE: We run this server-side!! Do not use this global except in the runs:can-run-more-tests routine
;;
(define *last-num-running-tests* 0)
;; (define *runs:can-run-more-tests-count* 0)
(define (runs:shrink-can-run-more-tests-count runsdat)
  (runs:dat-can-run-more-tests-count-set! runsdat 0))

(define (runs:inc-can-run-more-tests-count runsdat)
  (runs:dat-can-run-more-tests-count-set!
   runsdat
   (+ (runs:dat-can-run-more-tests-count runsdat) 1)))

;;  (set! *runs:can-run-more-tests-count* 0)) ;; (/ *runs:can-run-more-tests-count* 2)))

;; Temporary globals. Move these into the logic or into common
;;
(define *seen-cant-run-tests* (make-hash-table)) ;; use to track tests that we suspect cannot be run
(define (runs:inc-cant-run-tests testname)
  (hash-table-set! *seen-cant-run-tests* testname
		   (+ (hash-table-ref/default *seen-cant-run-tests* testname 0) 1)))

(define (runs:can-keep-running? testname n)
  (< (hash-table-ref/default *seen-cant-run-tests* testname 0) n))

(define *runs:denoise* (make-hash-table)) ;; key => last-time-ran

;; mechanism to limit printing info to the screen that is repetitive.
;;
;; Example: 
;; (if (runs:lownoise "waiting on tasks" 60)
;;     (debug:print-info 2 *default-log-port* "waiting for tasks to complete, sleeping briefly ..."))
;;
(define (runs:lownoise key waitval)

		     (if (member status '("TEN_STRIKES" "BLOCKED" "PREQ_FAIL" "ZERO_ITEMS" "PREQ_DISCARDED" "TIMED_OUT" ))
			 #f
			 t))
		    ((DELETED) #f)
		    (else t)))))
	  tests))

;; move all the miscellanea into this struct
;;
(defstruct runs:gendat inc-results inc-results-last-update inc-results-fmt run-info runname target)

(define *runs:general-data* 
  (make-runs:gendat
   inc-results: (make-hash-table)
   inc-results-last-update: 0
   inc-results-fmt: "~12a~12a~20a~12a~40a\n" ;; state status time duration test-name item-path
   run-info: #f
   runname: #f
   target: #f
   )
  )

(define (runs:incremental-print-results run-id)
  (let ((curr-sec    (current-seconds))
	(last-update (runs:gendat-inc-results-last-update *runs:general-data*)))
  (let* ((curr-sec    (current-seconds))
	 (runs-data   (bdat-runs-data *bdat*))
	 (last-update (runs:gendat-inc-results-last-update runs-data))
	 (runs-data   (bdat-runs-data *bdat*)))
    (if (> (- curr-sec last-update) 5) ;; at least five seconds since last update
	(let* ((run-dat  (or (runs:gendat-run-info *runs:general-data*)(rmt:get-run-info run-id)))
	       (runname  (or (runs:gendat-runname *runs:general-data*)
	(let* ((run-dat  (or (runs:gendat-run-info runs-data)(rmt:get-run-info run-id)))
	       (runname  (or (runs:gendat-runname runs-data)
			     (db:get-value-by-header (db:get-rows run-dat)
						     (db:get-header run-dat) "runname")))
	       (target   (or (runs:gendat-target *runs:general-data*)(rmt:get-target run-id)))
	       (target   (or (runs:gendat-target runs-data)(rmt:get-target run-id)))
	       (testsdat (let ((res (rmt:get-tests-for-run
				     run-id "%" '() '() ;; run-id testpatt states statuses
				     #f #f ;; offset limit
				     #f ;; not-in
				     #f ;; sort-by
				     #f ;; sort-order
				     #f ;; get full data (not 'shortlist)
				     last-update
				     'dashboard)))
			   (if (list? res)
			       res
			       (begin
				 (debug:print-error
				  0 *default-log-port*
				  "FAILED TO GET DATA using rmt:get-tests-for-run. Notify developers if you see this. result: " res)
				 '())))))
	  (runs:gendat-inc-results-last-update-set! *runs:general-data* (- curr-sec 1))
	  (if (not (runs:gendat-run-info *runs:general-data*))
	      (runs:gendat-run-info-set! *runs:general-data* run-dat))
	  (if (not (runs:gendat-runname  *runs:general-data*))
	      (runs:gendat-runname-set! *runs:general-data* runname))
	  (if (not (runs:gendat-target *runs:general-data*))
	      (runs:gendat-target-set! *runs:general-data* target))
	  (runs:gendat-inc-results-last-update-set! runs-data (- curr-sec 1))
	  (if (not (runs:gendat-run-info runs-data))
	      (runs:gendat-run-info-set! runs-data run-dat))
	  (if (not (runs:gendat-runname  runs-data))
	      (runs:gendat-runname-set! runs-data runname))
	  (if (not (runs:gendat-target runs-data))
	      (runs:gendat-target-set! runs-data target))
	  (for-each
	   (lambda (testdat)
	     (let* ((test-id    (db:test-get-id           testdat))
		    (prevdat    (hash-table-ref/default   (runs:gendat-inc-results *runs:general-data*)
		    (prevdat    (hash-table-ref/default   (runs:gendat-inc-results runs-data)
							  (conc run-id "," test-id) #f))
		    (test-name  (db:test-get-testname     testdat))
		    (item-path  (db:test-get-item-path    testdat))
		    (state      (db:test-get-state        testdat))
		    (status     (db:test-get-status       testdat))
		    (event-time (db:test-get-event_time   testdat))
		    (duration   (db:test-get-run_duration testdat)))
	       (if (and (not (member state '("DELETED" "REMOTEHOSTSTART" "RUNNING" "LAUNCHED""NOT_STARTED")))
			(not (and prevdat
				  (equal? state  (db:test-get-state  prevdat))
				  (equal? status (db:test-get-status prevdat)))))
		   (let ((fmt   (runs:gendat-inc-results-fmt *runs:general-data*))
		   (let ((fmt   (runs:gendat-inc-results-fmt runs-data))
			 (dtime (seconds->year-work-week/day-time event-time))) 
		     (if (runs:lownoise "inc-print" 600)
			 (format #t fmt "State" "Status" "Start Time" "Duration" "Test path"))
		     ;; (debug:print 0 *default-log-port* "fmt: " fmt " state: " state " status: " status " test-name: " test-name " item-path: " item-path " dtime: " dtime)
		     ;; (debug:print 0 #f "event-time: " event-time " duration: " duration)
		     (format #t fmt
			     state
			     status
			     dtime
			     (seconds->hr-min-sec duration)
			     (conc "lt/" target "/" runname "/" test-name (if (string-null? item-path) "" (conc "/" item-path))))
		     (hash-table-set! (runs:gendat-inc-results *runs:general-data*) (conc run-id "," test-id) testdat)))))
		     (hash-table-set! (runs:gendat-inc-results runs-data) (conc run-id "," test-id) testdat)))))
	   testsdat)))

    ;; I don't think this should be here? -- Matt
    #;(runs:gendat-inc-results-last-update-set! *runs:general-data* (- curr-sec 10))
    #;(runs:gendat-inc-results-last-update-set! runs-data (- curr-sec 10))

    ))

;; every time though the loop increment the test/itempatt val.
;; when the min is > max-allowed and none running then force exit
;;
(define *max-tries-hash* (make-hash-table))

(define (runs:pretty-long-list lst)
   (if (> (length lst) 8)(append (take lst 3)(list "...")) lst))

;;======================================================================
;; runs:run-tests-queue is called by runs:run-tests
;;======================================================================

	     (debug:print 5 *default-log-port* " exn=" (condition->list exn))
	     (thread-sleep! 1)
	     (tasks:open-db numretries (- numretries 1)))
	   (begin
	     (print-call-chain (current-error-port))
	     (debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
	     (debug:print 5 *default-log-port* " exn=" (condition->list exn))))
       (let* ((dbpath        (db:dbfile-path )) ;; (tasks:get-task-db-path))
       (let* ((dbpath        (common:get-db-tmp-area )) ;; (tasks:get-task-db-path))
	      (dbfile       (conc dbpath "/monitor.db"))
	      (avail        (tasks:wait-on-journal dbpath 10)) ;; wait up to about 10 seconds for the journal to go away
	      (exists       (common:file-exists? dbpath))
	      (write-access (file-writable? dbpath))
	      (mdb          (cond ;; what the hek is *toppath* doing here?
			     ((and (string? *toppath*)(file-writable? *toppath*))
			      (sqlite3:open-database dbfile))

    res))

;; 
#;(define (tasks:start-monitor db mdb)
  (if (> (tasks:get-num-alive-monitors mdb) 2) ;; have two running, no need for more
      (debug:print-info 1 *default-log-port* "Not starting monitor, already have more than two running")
      (let* ((megatestdb     (conc *toppath* "/megatest.db"))
	     (monitordbf     (conc (db:dbfile-path #f) "/monitor.db"))
	     (monitordbf     (conc (common:get-db-tmp-area) "/monitor.db"))
	     (last-db-update 0)) ;; (file-modification-time megatestdb)))
	(task:register-monitor mdb)
	(let loop ((count      0)
		   (next-touch 0)) ;; next-touch is the time where we need to update last_update
	  ;; if the db has been modified we'd best look at the task queue
	  (let ((modtime (file-modification-time megatestdbpath )))
	    (if (> modtime last-db-update)