Megatest

;; > 3 RUNNING with not test_dead do nothing (run should already be RUNNING/ na
;; > 0 RUNNING and test_dead then send KILLREQ ==> COMPLETED
;; 0 RUNNING ==> this is actually the first condition, should not get here

(define (runs:end-of-run-check run-id )
  (let* ((not-completed-cnt (rmt:get-not-completed-cnt run-id))  
	 (running-cnt (rmt:get-count-tests-running-for-run-id run-id))
	 (all-test-launched (rmt:get-var run-id (conc "launch-complete-" run-id)))
	 (current-state (rmt:get-run-state run-id))
	 (current-status (rmt:get-run-status run-id)))
    ;;get-vars run-id to query metadata table to check if all completed. if all-test-launched = yes then only not-completed-cnt = 0 means everyting is completed if no entry found in the table do nothing 
    (debug:print 0 *default-log-port* "Running test cnt :" running-cnt)                      
    (rmt:set-state-status-and-roll-up-run  run-id current-state current-status)
    (runs:update-junit-test-reporter-xml run-id) 
    (cond 

	  launchmod
	  processmod
	  rmtmod
	  runsmod
	  servermod
	  tasksmod
	  testsmod

	  ulex
	  )

;;   ;; ulex parameters
;;   (work-method 'direct)
;;   (return-method 'direct)
  
  ;; ulex parameters
  (work-method 'mailbox)
  (return-method 'mailbox)
  
;; fake out readline usage of toplevel-command
(define (toplevel-command . a) #f)
(define *didsomething* #f)  
(define *db* #f) ;; this is only for the repl, do not use in general!!!!

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

  (let* ((effective-toppath (or *toppath* apath)))
    (assert effective-toppath
	    "ERROR: get-pkts-dir called without *toppath* set. Exiting.")
    (let* ((pdir (conc effective-toppath "/.meta/srvpkts")))
      (if (file-exists? pdir)
	  pdir
	  (begin
	    (handle-exceptions ;; this exception handler should NOT be needed but ...
		exn
		pdir
	      (create-directory pdir #t))
	    pdir)))))

;; given a pkts dir read 
;;
(define (get-all-server-pkts pktsdir-in pktspec)
  (let* ((pktsdir  (if (file-exists? pktsdir-in)
		       pktsdir-in

    ;; Ensure all tests are registered in the test_meta table
    (runs:update-all-test_meta #f)

    ;; run the run prehook if there are no tests yet run for this run:
    ;;
    (runs:run-pre-hook run-id)
    ;; mark all test launched flag as false in the meta table 
    (rmt:set-var run-id (conc "launch-complete-" run-id) "no")
    (debug:print-info 1 *default-log-port* "Setting end-of-run to no")
    (let* ((config-reruns      (let ((x (configf:lookup *configdat* "setup" "reruns")))
			       (if x (string->number x) #f)))
	  (config-rerun-cnt (if config-reruns
			config-reruns
			1)))
    (if (eq? config-rerun-cnt run-count)

	  (debug:print-info 0 *default-log-port* "Have leftovers!")
	  (loop (car reg)(cdr reg) '() reruns))
	 (else
          (debug:print-info 4 *default-log-port* "cond branch - "  "rtq-9")
	  (debug:print-info 4 *default-log-port* "Exiting loop with...\n  hed=" hed "\n  tal=" tal "\n  reruns=" reruns))
	 ))) ;; end loop on sorted test names
    ;; this is the point where everything is launched and now you can mark the run in metadata table as all launched 
    (rmt:set-var run-id (conc "launch-complete-" run-id) "yes")  
        
    ;; now *if* -run-wait we wait for all tests to be done
    ;; Now wait for any RUNNING tests to complete (if in run-wait mode)
    ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat)))
    (thread-sleep! 10) ;; I think there is a race condition here. Let states/statuses settle
    
    (let wait-loop ((num-running      (rmt:get-count-tests-running-for-run-id run-id))

cleanup :
	killall mtest dboard -v -9 || true
	rm -rf *.log *.bak NB* logs/* .meta .db ../simpleruns/* lt

	 (l (string-length s)))
    (string->number (substring s (- l 3) l))
    ))

(define (run)
  (let* ((th1 (make-thread
	       (lambda ()
		 (let loop ((r 0)
			    (i 1)
			    (s 0)) ;; sum
		   (let ((start-time (current-milliseconds))
			 (run-id     (+ r (make-run-id))))
		     (rmt:register-test run-id "test1" (conc "item_" i))
		     (let* ((qry-time (- (current-milliseconds) start-time))
			    (tot-query-time (+ qry-time s))
			    (avg-query-time (* 1.0 (/ tot-query-time i))))
		       (if (> qry-time 500)
			   (print "WARNING: rmt:register-test took more than 500ms, "qry-time"ms, i="i", avg-query-time="avg-query-time))
		       (if (eq? (modulo i 100) 0)
			   (print "For run-id="run-id", "(rmt:get-keys-write)" num tests registered="i))
		       (if (< i 500)
			   (loop r (+ i 1) tot-query-time)
			   (if (< r 100)

			       (let* ((start-time (current-milliseconds)))
				 (print "rmt:get-keys "(rmt:get-keys)" in "(- (current-milliseconds) start-time))
				 (print "Got "(length (rmt:get-tests-for-run run-id "%" '() '() 0 #f #f #f #f #f 0 #f))" tests for run "run-id)
				 (print "Average query time: "avg-query-time)
				 (loop (+ r 1) 0 tot-query-time))))))))
	       )))
    (thread-start! th1)
    (thread-join! th1)))

)

(import junk)
(run)

# Valid values for state and status for steps, NB// It is not recommended you use this
[validvalues]
state start end completed

# Job tools are more advanced ways to control how your jobs are launched
[jobtools]
# useshell yes
launcher nbfake

# You can override environment variables for all your tests here
[env-override]
EXAMPLE_VAR example value

# As you run more tests you may need to add additional disks, the names are arbitrary but must be unique

(import srfi-18 
	test 
	chicken.string
	chicken.process-context
	chicken.file
	chicken.pretty-print
	commonmod
	ulex
	)

(define test-work-dir (current-directory))

(work-method   'mailbox)   ;; threads, direct, mailbox
(return-method 'mailbox)   ;; polling, mailbox, direct

;; given list of lists
;;  ( ( msg expected param1 param2 ...)
;;    ( ... ) )
;; apply test to all
;;
(define (test-batch proc pname inlst #!key (post-proc #f))

     
     ;; needed to get the interface:port that was automatically found
     udat-port
     udat-host-port
     
     ;; for testing only
     ;; pp-uconn
     
     ;; parameters
     work-method   ;; parameter; 'threads, 'mailbox, 'limited, 'direct
     return-method ;; parameter; 'mailbox, 'polling, 'direct
     )

(import scheme
	chicken.base
	chicken.file
	chicken.time
	chicken.condition

  (work-queue-thread #f)
  (num-threads-running 0)
  ) 

;; Parameters

;; work-method:
(define work-method (make-parameter 'direct))
;;    mailbox - all rdat goes through mailbox
;;    threads - all rdat immediately executed in new thread


;;    direct  - no queuing
;;

;; return-method, return the result to waiting send-receive:
(define return-method (make-parameter 'direct))
;;    mailbox - create a mailbox and use it for passing returning results to send-receive
;;    polling - put the result in a hash table keyed by qrykey and send-receive can poll it for result
;;    direct  - no queuing, result is passed back in single tcp connection
;;

;; ;; struct for keeping track of others we are talking to
;; ;;
;; (defstruct pdat
;;   (host-port  #f)
;;   (conns      '()) ;; list of pcon structs, pop one off when calling the peer
;;   )
;; 

;;        - I believe (without substantial evidence) that re-using connections will
;;          be beneficial ...
;;
(define (send udata host-port qrykey cmd params)
  (let* ((my-host-port (udat-host-port udata))          ;; remote will return to this
	 (isme         #f #;(equal? host-port my-host-port)) ;; calling myself?
	 ;; dat is a self-contained work block that can be sent or handled locally
	 (dat          (list my-host-port qrykey cmd params)))

    (cond
     (isme (ulex-handler udata dat)) ;; no transmission needed
     (else
      (handle-exceptions ;; TODO - MAKE THIS EXCEPTION CMD SPECIFIC?
	  exn
	  #f
	(begin
	  ;; (mutex-lock! *send-mutex*)
	  (let-values (((inp oup)(tcp-connect host-port)))
	    (let ((res (if (and inp oup)
			   (begin
			     (serialize dat oup)
			     (deserialize inp))
			   (begin
			     (print "ERROR: send called but no receiver has been setup. Please call setup first!")
			     #f))))
	      (close-input-port inp)
	      (close-output-port oup)
	      ;; (mutex-unlock! *send-mutex*)
	      res)))))))) ;; res will always be 'ack unless return-method is direct
  
;; send a request to the given host-port and register a mailbox in udata
;; wait for the mailbox data and return it
;;
(define (send-receive uconn host-port cmd data)
  (cond
   ((member cmd '(ping goodbye)) ;; these are immediate
    (send uconn host-port 'ping cmd data))
   ((eq? (work-method) 'direct)
    ;; the result from send will be the actual result, not an 'ack
    (send uconn host-port 'direct cmd data))
   (else
    (case (return-method)
      ((polling)
       (let* ((qrykey (make-cookie uconn))
	      (sres   (send uconn host-port qrykey cmd data)))
	 (case sres
	   ((ack)
	    (let loop ((start-time (current-milliseconds)))
	      (if (> (current-milliseconds)(+ start-time 10000)) ;; ten seconds timeout
		  (begin
		    (print "ULEX ERROR: timed out waiting for response from "host-port", "cmd" "data)
		    #f)
		  (let* ((result (hash-table-ref/default (udat-mboxes uconn) qrykey #f))) ;; NOTE: we are re-using mboxes hash
		    (if result ;; result is '(status . result-data) or #f for nothing yet
			(begin
			  (hash-table-delete! (udat-mboxes uconn) qrykey)
			  (cdr result))
			(begin
			  (thread-sleep! 0.01)
			  (loop start-time)))))))
	   (else
	    (print "ULEX ERROR: Communication failed? sres="sres)
	    #f))))
      ((mailbox) 
       (let* ((cmbox     (get-cmbox uconn)) ;; would it be better to keep a stack of mboxes to reuse?
	      (qrykey    (car cmbox))
	      (mbox      (cdr cmbox))
	      (mbox-time (current-milliseconds))
	      (sres      (send uconn host-port qrykey cmd data))) ;; short res
	 (if (eq? sres 'ack)
	     (let* ((mbox-timeout-secs    120 #;(if (eq? 'primordial (thread-name (current-thread)))
					  #f
					  120)) ;; timeout)
		    (mbox-timeout-result 'MBOX_TIMEOUT)
		    (res                  (mailbox-receive! mbox mbox-timeout-secs mbox-timeout-result))
		    (mbox-receive-time    (current-milliseconds)))
	       ;; (put-cmbox uconn cmbox) ;; reuse mbox and cookie. is it worth it?
	       (hash-table-delete! (udat-mboxes uconn) qrykey)
	       (if (eq? res 'MBOX_TIMEOUT)
		   (begin
		     (print "WARNING: mbox timed out for query "cmd", with data "data", waiting for response from "host-port".")

		     ;; here it might make sense to clean up connection records and force clean start?
		     ;; NO. The progam using ulex needs to do the reset. Right thing here is exception
		     
		     #f)  ;; convert to raising exception?
		   res))
	     (begin
	       (print "ERROR: Communication failed? Got "sres)
	       #f))))
      (else
       (print "ULEX ERROR: unrecognised return-method "(return-method)".")
       #f)))))

;;======================================================================
;; responder side
;;======================================================================

;; take a request, rdat, and if not immediate put it in the work queue
;;
;; Reserved cmds; ack ping goodbye response
;;
(define (ulex-handler uconn rdat)
  (assert (list? rdat) "FATAL: ulex-handler give rdat as not list")
  (match rdat ;;  (string-split controldat)
    ((rem-host-port qrykey cmd params)
     ;; (print "ulex-handler got: "rem-host-port" qrykey: "qrykey" cmd: "cmd" params: "params)

     (case cmd
       ;; ((ack )(print "Got ack! But why? Should NOT get here.") 'ack)
       ((ping)
	;; (print "Got Ping!")
	;; (add-to-work-queue uconn rdat)
	'ack)
       ((goodbye)
	;; just clear out references to the caller. NOT COMPLETE
	(add-to-work-queue uconn rdat)
	'ack)
       ((response) ;; this is a result from remote processing, send it as mail ...
	(case (return-method)
	  ((polling)
	   (hash-table-set! (udat-mboxes uconn) qrykey (cons 'ok params))
	   'ack)
	  ((mailbox)
	   (let ((mbox (hash-table-ref/default (udat-mboxes uconn) qrykey #f)))
	     (if mbox
		 (begin
		   (mailbox-send! mbox params) ;; params here is our result
		   'ack)
		 (begin
		   (print "ERROR: received result but no associated mbox for cookie "qrykey)
		   'no-mbox-found))))
	  (else (print "ULEX ERROR: unrecognised return-method "(return-method))
		'bad-return-method)))
       (else ;; generic request - hand it to the work queue
	(add-to-work-queue uconn rdat)
	'ack)))
    (else
     (print "ULEX ERROR: bad rdat "rdat)

     'bad-rdat)))

;; given an already set up uconn start the cmd-loop
;;
(define (ulex-cmd-loop uconn)
  (let* ((serv-listener (udat-socket uconn))
	 (listener      (lambda ()
			  (let loop ((state 'start))

     (mailbox-send! (udat-work-queue uconn) rdat))
    ((direct)
     (do-work uconn rdat))
    (else
     (print "ULEX ERROR: work-method "(work-method)" not recognised, using mailbox.")
     (mailbox-send! (udat-work-queue uconn) rdat))))
     
;; move the logic to return the result somewhere else?
;;
(define (do-work uconn rdat)
  (let* ((proc (udat-work-proc uconn))) ;; get it each time - conceivebly it could change
    ;; put this following into a do-work procedure
    (match rdat
      ((rem-host-port qrykey cmd params)
       (let* ((start-time (current-milliseconds))
	      (result (proc rem-host-port qrykey cmd params))
	      (end-time (current-milliseconds))
	      (run-time (- end-time start-time)))
	 (case (work-method)
	   ((direct) result)
	   (else
	    (print "ULEX: work "cmd", "params" done in "run-time" ms")
	    ;; send 'response as cmd and result as params
	    (send uconn rem-host-port qrykey 'response result) ;; could check for ack
	    (print "ULEX: response sent back to "rem-host-port" in "(- (current-milliseconds) end-time))))))
      (MBOX_TIMEOUT #f)
      (else
       (print "ERROR: rdat "rdat", did not match rem-host-port qrykey cmd params")))))

;; NEW APPROACH:
;;   
(define (process-work-queue uconn)