Megatest

(define *wdnum* 0)
(define *wdnum*mutex (make-mutex))
;; currently the primary job of the watchdog is to run the sync back to megatest.db from the db in /tmp
;; if we are on the homehost and we are a server (by definition we are on the homehost if we are a server)
;;


(define (common:readonly-watchdog dbstruct)
  (thread-sleep! 0.05) ;; delay for startup

  ;; sync megatest.db to /tmp/.../megatst.db
  (let ((sync-cool-off-duration   3)
        (golden-mtdb     (dbr:dbstruct-mtdb dbstruct))
        (golden-mtpath   (db:dbdat-get-path mtdb))
        (tmp-mtdb        (dbr:dbstruct-tmpdb dbstruct))
        (tmp-mtpath      (db:dbdat-get-path mtdb)))
    (debug:print-info 0 *default-log-port* "Read-only periodic sync thread started.")
    (let loop ((last-sync-time 0))
      (let* ((duration-since-last-sync (- (current-seconds) last-sync-time)))
        (if (and (not *time-to-exit*)
                 (< duration-since-last-sync sync-cool-off-duration))
            (thread-sleep! (- sync-cool-off-duration duration-since-last-sync)))
        (if (not *time-to-exit*)
            (let ((golden-mtdb-mtime (file-modification-time golden-mtpath))
                  (tmp-mtdb-mtime    (file-modification-time tmp-mtpath)))
              (if (> golden-mtdb-mtime tmp-mtdb-mtime)
                  (let ((res (db:multi-db-sync dbstruct 'old2new)))
                    (debug:print-info 0 *default-log-port* "rosync called, " res " records transferred."))
                  (loop (current-seconds)))
              #t))))
    (debug:print-info 0 *default-log-port* "Exiting readonly-watchdog timer, *time-to-exit* = " *time-to-exit*" pid="(current-process-id)" mtpath="golden-mtpath)))


        
(define (common:watchdog)
(define (common:writable-watchdog dbstruct)
  (thread-sleep! 0.05) ;; delay for startup
  (let ((legacy-sync (common:run-sync?))
	(debug-mode  (debug:debug-mode 1))
	(last-time   (current-seconds))
        (this-wd-num     (begin (mutex-lock! *wdnum*mutex) (let ((x *wdnum*)) (set! *wdnum* (add1 *wdnum*)) (mutex-unlock! *wdnum*mutex) x))))
    (debug:print-info 3 *default-log-port* "watchdog starting. legacy-sync is " legacy-sync" pid="(current-process-id)" this-wd-num="this-wd-num)
    (if (and legacy-sync (not *time-to-exit*))
	(let* ((dbstruct (db:setup))
	(let* (;;(dbstruct (db:setup))
	       (mtdb     (dbr:dbstruct-mtdb dbstruct))
	       (mtpath   (db:dbdat-get-path mtdb)))
	  (debug:print-info 0 *default-log-port* "Server running, periodic sync started.")
	  (let loop ()
	    ;; sync for filesystem local db writes
	    ;;
	    (mutex-lock! *db-multi-sync-mutex*)

			   (< count 4)) ;; was 11, changing to 4. 
		      (begin
			(thread-sleep! 1)
			(delay-loop (+ count 1))))
		  (if (not *time-to-exit*) (loop))))
	    (if (common:low-noise-print 30)
		(debug:print-info 0 *default-log-port* "Exiting watchdog timer, *time-to-exit* = " *time-to-exit*" pid="(current-process-id)" this-wd-num="this-wd-num)))))))

;; TODO: for multiple areas, we will have multiple watchdogs; and multiple threads to manage
(define (common:watchdog)
  (let ((dbstruct (db:setup)))
    (if (dbstruct-readonly dbstruct)
        (common:readonly-watchdog dbstruct)
        (common:writable-watchdog dbstruct))))


(define (std-exit-procedure)
  (on-exit (lambda () 0))
  ;;(BB> "std-exit-procedure called; *time-to-exit*="*time-to-exit*)
  (let ((no-hurry  (if *time-to-exit* ;; hurry up
		       #f
		       (begin

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

;; each db entry is a pair ( db . dbfilepath )
;; I propose this record evolves into the area record
;;
(defstruct dbr:dbstruct 
  ;; (tmpdb       #f)
  (tmpdb       #f)
  (dbstack     #f) ;; stack for tmp db handles, do not initialize with a stack
  (mtdb        #f)
  (refndb      #f)
  (homehost    #f) ;; not used yet
  (on-homehost #f) ;; not used yet
  (read-only   #f)
  )                ;; goal is to converge on one struct for an area but for now it is too confusing
  

;; record for keeping state,status and count for doing roll-ups in
;; iterated tests
;;
(defstruct dbr:counts

;;     (dbr:dbstruct-olddb-set!  dbstruct olddb)
;;     ;;; (mutex-unlock! *rundb-mutex*) ;;; why did we need a mutex on opening db's?
;;     (db:sync-tables db:sync-tests-only *megatest-db* db)
;;     db))

;; 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))
(define (db:open-db dbstruct #!key (areapath #f)) ;; 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* ((dbpath       (db:dbfile-path)) ;;  0))
               (dbexists     (file-exists? dbpath))
	       (dbfexists    (file-exists? (conc dbpath "/megatest.db")))
               (tmpdb        (db:open-megatest-db path: dbpath)) ;; lock-create-open dbpath db:initialize-main-db))
               (mtdb         (db:open-megatest-db))
               (refndb       (db:open-megatest-db path: dbpath name: "megatest_ref.db"))
               (write-access (file-write-access? dbpath)))
          (if (and dbexists (not write-access))
              (set! *db-write-access* #f))
              (begin (set! *db-write-access* #f)
                     (dbr:dbstruct-readonly-set! dbstruct #t)))
          (dbr:dbstruct-mtdb-set!   dbstruct mtdb)
          (dbr:dbstruct-tmpdb-set!  dbstruct tmpdb)
          (dbr:dbstruct-dbstack-set! dbstruct (make-stack))
          (dbr:dbstruct-dbstack-set! dbstruct (make-stack)) ;; BB: why a stack?  Why would the number of db's be indeterminate?  Is this a legacy of 1.db 2.db .. ?
          (stack-push! (dbr:dbstruct-dbstack dbstruct) tmpdb) ;; olddb is already a (cons db path)
          (dbr:dbstruct-refndb-set! dbstruct refndb)
          ;;	    (mutex-unlock! *rundb-mutex*)
          (if (and (not dbfexists)
                   write-access) ;; *db-write-access*) ;; did not have a prior db and do have write access
	      (begin
		(debug:print 0 *default-log-port* "filling db " (db:dbdat-get-path tmpdb) " with data from " (db:dbdat-get-path mtdb))
		(db:sync-tables (db:sync-all-tables-list dbstruct) #f mtdb refndb tmpdb))
	      (debug:print 0 *default-log-port* " db, " (db:dbdat-get-path tmpdb) " already exists, not propogating data from " (db:dbdat-get-path mtdb)))
	  ;; (db:multi-db-sync dbstruct 'old2new))  ;; migrate data from megatest.db automatically
          tmpdb))))

;; Make the dbstruct, setup up auxillary db's and call for main db at least once
;;
;; called in http-transport and replicated in rmt.scm for *local* access. 
;;
(define (db:setup #!key (areapath #f))
  (or *dbstruct-db*
  (or *dbstruct-db* ;; TODO: when multiple areas are supported, this optimization will be a hazard
      (if (common:on-homehost?)
	  (let* ((dbstruct (make-dbr:dbstruct)))
	    (db:open-db dbstruct areapath: areapath)
	    (set! *dbstruct-db* dbstruct)
	    dbstruct)
	  (begin
	    (debug:print 0 *default-log-port* "ERROR: attempt to open database when not on homehost. Exiting. Homehost: " (common:get-homehost))

   (cond
    ((not fromdb) (debug:print 3 *default-log-port* "WARNING: db:sync-tables called with fromdb missing") -1)
    ((not todb)   (debug:print 3 *default-log-port* "WARNING: db:sync-tables called with todb missing") -2)
    ((not (sqlite3:database? (db:dbdat-get-db fromdb)))
     (debug:print-error 0 *default-log-port* "db:sync-tables called with fromdb not a database " fromdb) -3)
    ((not (sqlite3:database? (db:dbdat-get-db todb)))
     (debug:print-error 0 *default-log-port* "db:sync-tables called with todb not a database " todb) -4)

    ((not (file-write-access? (cadr todb)))
     (debug:print-error 0 *default-log-port* "db:sync-tables called with todb not a read-only database " todb) -5)
    ((not (null? (let ((readonly-slave-dbs
                        (filter
                         (lambda (dbdat)
                           (not (file-write-access? (cadr todb))))
                         
                         slave-dbs)))
                   (for-each
                    (lambda (bad-dbdat)
                      (debug:print-error
                       0 *default-log-port* "db:sync-tables called with todb not a read-only database " bad-dbdat))
                    readonly-slave-dbs)
                   readonly-slave-dbs))) -6)
    (else
     (let ((stmts       (make-hash-table)) ;; table-field => stmt
	   (all-stmts   '())              ;; ( ( stmt1 value1 ) ( stml2 value2 ))
	   (numrecs     (make-hash-table))
	   (start-time  (current-milliseconds))
	   (tot-count   0))
       (for-each ;; table

	  ))))

;; options:
;;
;;  'killservers  - kills all servers
;;  'dejunk       - removes junk records
;;  'adj-testids  - move test-ids into correct ranges
;;  'old2new      - sync megatest.db records to .db/{main,1,2 ...}.db
;;  'new2old      - sync .db/{main,1,2,3 ...}.db to megatest.db
;;  'old2new      - sync megatest.db to /tmp/.../megatest.db and /tmp/.../megatest_ref.db
;;  'new2old      - sync /tmp/.../megatest.db to megatest.db and /tmp/.../megatest_ref.db (and update data_synced)
;;  'closeall     - close all opened dbs
;;  'schema       - attempt to apply schema changes
;;
;;  run-ids: '(1 2 3 ...) or #f (for all)
;;
(define (db:multi-db-sync dbstruct . options)
  (if (not (launch:setup))
      (debug:print 0 *default-log-port* "ERROR: not able to setup up for megatest.")
      (let* ((mtdb     (dbr:dbstruct-mtdb dbstruct))
	     (tmpdb    (db:get-db dbstruct))

	;;       (db:delay-if-busy mtdb)
	;;       (db:prep-megatest.db-for-migration mtdb)))

	;; sync runs, test_meta etc.
	;;
	(if (member 'old2new options)
	    ;; (begin
            (set! data-synced
	    (db:sync-tables (db:sync-all-tables-list dbstruct) #f mtdb tmpdb refndb))
                  (+ (db:sync-tables (db:sync-all-tables-list dbstruct) #f mtdb tmpdb refndb)
                     data-synced)))
			      ;; (db:sync-main-list mtdb) mtdb (db:get-db dbstruct #f))
;; 	      (for-each 
;; 	       (lambda (run-id)
;; 		 (db:delay-if-busy mtdb)
;; 		 (let ((testrecs (db:get-all-tests-info-by-run-id mtdb run-id)))
;; ;;		       (dbstruct (if toppath (make-dbr:dbstruct path: toppath local: #t) #f)))
;; 		   (debug:print 0 *default-log-port* "INFO: Propagating " (length testrecs) " records for run-id=" run-id " to run specific db")
;; 		   (db:replace-test-records dbstruct run-id testrecs)
;; 		   (sqlite3:finalize! (db:dbdat-get-db (dbr:dbstruct-rundb dbstruct)))))
;; 	       run-ids)))

	;; now ensure all newdb data are synced to megatest.db
	;; do not use the run-ids list passed in to the function
	;;
	(if (member 'new2old options)
	    (set! data-synced
		  (+ (db:sync-tables (db:sync-all-tables-list dbstruct) #f tmpdb refndb mtdb)
		      data-synced)))



        (if (member 'fixschema options)
            (begin
              (db:patch-schema-maindb (db:dbdat-get-db mtdb))
              (db:patch-schema-maindb (db:dbdat-get-db tmpdb))
              (db:patch-schema-maindb (db:dbdat-get-db refndb))

;; immediately set MT_TARGET if -reqtarg or -target are available
;;
(let ((targ (or (args:get-arg "-reqtarg")(args:get-arg "-target"))))
  (if targ (setenv "MT_TARGET" targ)))

;; The watchdog is to keep an eye on things like db sync etc.
;;

;; TODO: for multiple areas, we will have multiple watchdogs; and multiple threads to manage
(define *watchdog* (make-thread common:watchdog "Watchdog thread"))

(if (not (args:get-arg "-server"))
    (thread-start! *watchdog*)) ;; if starting a server; wait till we get to running state before kicking off watchdog

;; bracket open-output-file with code to make leading directory if it does not exist and handle exceptions
(define (open-logfile logpath)

(if (not *didsomething*)
    (debug:print 0 *default-log-port* help))
;;(BB> "thread-join! watchdog")

;; join the watchdog thread if it has been thread-start!ed  (it may not have been started in the case of a server that never enters running state)
;;   (symbols returned by thread-state: created ready running blocked suspended sleeping terminated dead)
;; TODO: for multiple areas, we will have multiple watchdogs; and multiple threads to manage
(if (thread? *watchdog*)
    (case (thread-state *watchdog*)
      ((ready running blocked sleeping terminated dead)
       (thread-join! *watchdog*))))

(set! *time-to-exit* #t)

  (mutex-lock! *rmt-mutex*)

  ;; 1. check if server is started IFF cmd is a write OR if we are not on the homehost, store in runremote
  ;; 2. check the age of the connections. refresh the connection if it is older than timeout-20 seconds.
  ;; 3. do the query, if on homehost use local access
  ;;
  (let* ((start-time (current-seconds)) ;; snapshot time so all use cases get same value
         (areapath *toppath*);; TODO - resolve from dbstruct to be compatible with multiple areas
         (readonly-mode (file-write-access? (conc *toppath* "/megatest.db")  )) ;; TODO: use dbstruct or runremote to figure this out in future
	 (runremote  (or area-dat *runremote*)))
    (cond
     ;; give up if more than 15 attempts
     ((> attemptnum 15)
      (debug:print 0 *default-log-port* "ERROR: 15 tries to start/connect to server. Giving up.")
      (exit 1))

     ;; readonly mode, read request-  handle it - case 20
     ((and readonly-mode
           (member cmd api:read-only-queries)) 
      (mutex-unlock! *rmt-mutex*)
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case 20")
      (rmt:open-qry-close-locally cmd 0 params)
      )

     ;; readonly mode, write request.  Do nothing, return #f
     (readonly-mode
      (mutex-unlock! *rmt-mutex*)
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case 21")
      (debug:print 0 *default-log-port* "WARNING: write transaction requested on a readonly area.  cmd="cmd" params="params)
      #f
      )

     ;; reset the connection if it has been unused too long
     ((and runremote
           (remote-conndat runremote)
	   (let ((expire-time (+ (- start-time (remote-server-timeout runremote))(random 30)))) ;; add 30 seconds of noise so that not all running tests expire at the same time causing a storm of server starts
	     (< (http-transport:server-dat-get-last-access (remote-conndat runremote)) expire-time)))
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case  8")
      (remote-conndat-set! runremote #f)