Megatest

;; 	      (car load-res))
;;     cpu-load))

;; get values from cached info from dropping file in logs dir
;;  e.g. key is host and dtype is normalized-load
;;
(define (common:get-cached-info key dtype #!key (age 5))
  (let* ((fullpath (conc *toppath* "/.sysdata/" key "-" dtype ".log")))
    (if (and (file-exists? fullpath)
	     (file-read-access? fullpath))
	(handle-exceptions
	 exn
	 #f
	 (debug:print 2 *default-log-port* "reading file " fullpath)
	 (let ((real-age (- (current-seconds)(file-change-time fullpath)))) 
	   (if (< real-age age)
	       (with-input-from-file fullpath read)
	       (begin
		 (debug:print 1 *default-log-port* "file " fullpath " is too old (" real-age" seconds) to trust, skipping reading it")
		 #f))))
	(begin
	  (debug:print 2 *default-log-port* "not reading file " fullpath)
	  #f))))
 
(define (common:write-cached-info key dtype dat)
  (let* ((fulldir  (conc *toppath* "/.sysdata"))
	 (fullpath (conc fulldir "/" key "-" dtype ".log")))
    (if (not (file-exists? fulldir))(create-directory fulldir #t))
    (handle-exceptions
     exn
     #f
     (with-output-to-file fullpath (lambda ()(pp dat))))))

;; get cpu load by reading from /proc/loadavg, return all three values
;;

     (else default))))

(define (common:get-normalized-cpu-load-raw remote-host)
  (let* ((actual-host (or remote-host (get-host-name)))) ;; #f is localhost
    (or (common:get-cached-info actual-host "normalized-load")
	(let ((data (if remote-host
			(with-input-from-pipe 
			    (conc "ssh " remote-host " \"cat /proc/loadavg;cat /proc/cpuinfo;echo end\"")
			  read-lines)
			(append 
			 (with-input-from-file "/proc/loadavg" 
			   read-lines)
			 (with-input-from-file "/proc/cpuinfo"
			   read-lines)
			 (list "end"))))

	      #f
	      (let loop ((hed      (car data))
			 (tal      (cdr data))
			 (loads    #f)
			 (proc-num 0)  ;; processor includes threads
			 (phys-num 0)  ;; physical chip on motherboard
			 (core-num 0)) ;; core
		;;; (print hed ", " loads ", " proc-num ", " phys-num ", " core-num)
		(if (null? tal) ;; have all our data, calculate normalized load and return result
		    (let* ((act-proc (+ proc-num 1))
			   (act-phys (+ phys-num 1))
			   (act-core (+ core-num 1))
			   (adj-proc-load (/ (car loads) act-proc))
			   (adj-core-load (/ (car loads) act-core))
			   (result

	      (if new-best
		  (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 maxload msg)
  (let* ((hh-dat (if (common:on-homehost?) ;; if we are on the homehost then pass in #f so the calls are local.
                     #f
                     (common:get-homehost)))
         (hh     (if hh-dat (car hh-dat) #f)))

    (common:wait-for-normalized-load maxload msg hh)))

(define (common:get-num-cpus remote-host)
  (let* ((actual-host (or remote-host (get-host-name))))
    (or (common:get-cached-info actual-host "num-cpus" age: (+ 2592000 (random 3600))) ;; hosts had better not be changing the number of cpus too often!
	(let* ((proc   (lambda ()
			 (let loop ((numcpu 0)
				    (inl    (read-line)))
			   (if (eof-object? inl)
			       (begin
				 (common:write-cached-info actual-host "num-cpus" numcpu)
				 numcpu)
			       (loop (if (string-match "^processor\\s+:\\s+\\d+$" inl)
					 (+ numcpu 1)
					 numcpu)
				     (read-line))))))
	       (result (if remote-host
			   (with-input-from-pipe 
			       (conc "ssh " remote-host " cat /proc/cpuinfo")
			     proc)
			   (with-input-from-file "/proc/cpuinfo" proc))))
	  (common:write-cached-info actual-host "num-cpus" result)
	  result))))

;; wait for normalized cpu load to drop below maxload
;;
(define (common:wait-for-normalized-load maxload msg remote-host)
  (let ((num-cpus (common:get-num-cpus remote-host)))
    (common:wait-for-cpuload maxload num-cpus 15 msg: msg remote-host: remote-host)))

;; DO NOT CALL THIS DIRECTLY. It is called from common:wait-for-normalized-load
;;
(define (common:wait-for-cpuload maxload-in numcpus-in waitdelay #!key (count 1000) (msg #f)(remote-host #f)(force-maxload #f))
  (let* ((loadavg (common:get-cpu-load remote-host))
	 (numcpus (if (<= 1 numcpus-in) ;; not possible to have zero.  If we get 1, it's possible that we got the previous default, and we should check again
		      (common:get-num-cpus remote-host)
		      numcpus-in))
	 (maxload (if force-maxload
		      maxload-in
		      (max maxload-in 0.5))) ;; so maxload must be greater than 0.5 for now BUG - FIXME?
	 (first   (car loadavg))
	 (next    (cadr loadavg))
	 (adjload (* maxload (max 1 numcpus))) ;; possible bug where numcpus (or could be maxload) is zero, crude fallback is to at least use 1
	 (loadjmp (- first next))
         (adjwait (min (+ 300 (random 10)) (abs (* (+ (random 10)(/ (- 1000 count) 10) waitdelay) (- first adjload) ))  )));; add some randomness to the time to break any alignment where netbatch dumps many jobs to machines simultaneously
    (debug:print-info 1 *default-log-port* "Checking cpuload on " (or remote-host "localhost") ", maxload: " maxload
		      ", load: " first ", adjload: " adjload ", loadjmp: " loadjmp)
    (cond
     ((and (> first adjload)
	   (> count 0))
      (debug:print-info 0 *default-log-port* "server start delayed " adjwait " seconds due to load " first " exceeding max of " adjload " on server " (or remote-host (get-host-name)) " (normalized load-limit: " maxload ") " (if msg msg ""))
      (thread-sleep! adjwait)
      (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: remote-host))
     ((and (> loadjmp numcpus)
	   (> count 0))
      (debug:print-info 0 *default-log-port* "waiting " adjwait " seconds due to load jump " loadjmp " > numcpus " numcpus (if msg msg ""))
      (thread-sleep! adjwait)
      (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: 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))))

;;     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 1.6543)

(use format typed-records) ;; RADT => purpose of json format??

(declare (unit rmt))
(declare (uses api))
(declare (uses http-transport))
(include "common_records.scm")
;; (declare (uses rmtmod))

;; (import rmtmod)

;;
;; THESE ARE ALL CALLED ON THE CLIENT SIDE!!!
;;

;; generate entries for ~/.megatestrc with the following
;;

(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)))
                          
  (if (> attemptnum 2)
      (debug:print 0 *default-log-port* "INFO: attemptnum in rmt:send-receive is " attemptnum))
    
  (cond
   ((> attemptnum 2) (thread-sleep! 0.05))
   ((> attemptnum 10) (thread-sleep! 0.5))
   ((> attemptnum 20) (thread-sleep! 1)))
  (if (and (> attemptnum 5) (= 0 (modulo attemptnum 15)))  
    (begin (rmt:start-server rid) (thread-sleep! 3))) 
  
  
  ;;DOT digraph megatest_state_status {
  ;;DOT   ranksep=0;
  ;;DOT   // rankdir=LR;
  ;;DOT   node [shape="box"];
  ;;DOT "rmt:send-receive" -> MUTEXLOCK;
  ;;DOT { edge [style=invis];"case 1" -> "case 2" -> "case 3" -> "case 4" -> "case 5" -> "case 6" -> "case 7" -> "case 8" -> "case 9" -> "case 10" -> "case 11"; }
  ;; do all the prep locked under the rmt-mutex
  (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
	 (runremote     (or area-dat
			    *runremote*))
         (attemptnum    (+ 1 attemptnum))
	 (readonly-mode (rmtmod:calc-ro-mode runremote *toppath*)))

    ;; DOT INIT_RUNREMOTE; // leaving off - doesn't really add to the clarity
    ;; DOT MUTEXLOCK -> INIT_RUNREMOTE [label="no remote?"];
    ;; DOT INIT_RUNREMOTE -> MUTEXLOCK;
    ;; ensure we have a record for our connection for given area
    (if (not runremote)                   ;; can remove this one. should never get here.         

	(thread-sleep! 0.1) ;; since we shouldn't get here, delay a little
	(remote-hh-dat-set! runremote (common:get-homehost)))
    
    ;;(print "BB> readonly-mode is "readonly-mode" dbfile is "dbfile)
    (cond
     ;;DOT EXIT;
     ;;DOT MUTEXLOCK -> EXIT [label="> 15 attempts"]; {rank=same "case 1" "EXIT" }
     ;; give up if more than 150 attempts
     ((> attemptnum 150)
      (debug:print 0 *default-log-port* "ERROR: 150 tries to start/connect to server. Giving up.")
      (exit 1))

     ;;DOT CASE2 [label="local\nreadonly\nquery"];
     ;;DOT MUTEXLOCK -> CASE2; {rank=same "case 2" CASE2}
     ;;DOT CASE2 -> "rmt:open-qry-close-locally";
     ;; readonly mode, read request-  handle it - case 2
     ((and readonly-mode

(define (rmt:test-set-archive-block-id run-id test-id archive-block-id)
  (rmt:send-receive 'test-set-archive-block-id run-id (list run-id test-id archive-block-id)))

(define (rmt:test-get-archive-block-info archive-block-id)
  (rmt:send-receive 'test-get-archive-block-info #f (list archive-block-id)))


(define (rmtmod:calc-ro-mode runremote *toppath*)
  (if (and runremote
	   (remote-ro-mode-checked runremote))
      (remote-ro-mode runremote)
      (let* ((dbfile  (conc *toppath* "/megatest.db"))
	     (ro-mode (not (file-write-access? dbfile)))) ;; TODO: use dbstruct or runremote to figure this out in future
	(if runremote
	    (begin
	      (remote-ro-mode-set! runremote ro-mode)
	      (remote-ro-mode-checked-set! runremote #t)
	      ro-mode)
	    ro-mode))))

(define (extras-readonly-mode rmt-mutex log-port cmd params)
  (mutex-unlock! rmt-mutex)
  (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*)
  (remote-conndat-set!    runremote #f)
  (http-transport:close-connections area-dat: 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)
	   (eq? (vector-length res) 2)
	   (eq? (vector-ref res 1) 'overloaded)) ;; since we are
						 ;; looking at the
						 ;; data to carry the
						 ;; error we'll use a
						 ;; fairly obtuse
						 ;; combo to minimise
						 ;; the chances of
						 ;; some sort of
						 ;; collision.  this
						 ;; is the case where
						 ;; the returned data
						 ;; is bad or the
						 ;; 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 area-dat: 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)

  (set! debug:print                      dbgp)
  (set! debug:print-info                 dbgpinfo)
  (set! remote-ro-mode                   ro-mode)
  (set! remote-ro-mode-set!              ro-mode-set)
  (set! remote-ro-mode-checked-set!      ro-mode-checked-set)
  (set! remote-ro-mode-checked           ro-mode-checked))



























































)

					(debug:print-error 0 *default-log-port* "prereqs-not-met is not a list! " prereqs-not-met)
					'())))
	 (non-completed           (filter (lambda (x)             ;; remove hed from not completed list, duh, of course it is not completed!
					    (not (equal? x hed)))
					  (runs:calc-not-completed prereqs-not-met)))
	 (loop-list               (list hed tal reg reruns))
	 ;; configure the load runner

	 (maxload                 (string->number (or (configf:lookup *configdat* "jobtools" "maxload") "3.0")))         ;; use a non-number string to disable
         (maxhomehostload         (string->number (or (configf:lookup *configdat* "jobtools" "maxhomehostload") "2.0"))) ;; use a non-number string to disable
         (waitdelay               (string->number (or (configf:lookup *configdat* "jobtools" "waitdelay") "60"))))
    (debug:print-info 4 *default-log-port* "have-resources: " have-resources " prereqs-not-met: (" 
		      (string-intersperse 
		       (map (lambda (t)
			      (if (vector? t)

	       (and (member 'toplevel testmode) ;;  'toplevel)
		    (null? non-completed)
		    (not (member 'exclusive testmode)))))
      ;; (hash-table-delete! *max-tries-hash* (db:test-make-full-name test-name item-path))
      ;; we are going to reset all the counters for test retries by setting a new hash table
      ;; this means they will increment only when nothing can be run
      (set! *max-tries-hash* (make-hash-table))
      
      ;; well, first lets see if cpu load throttling is enabled. If so wait around until the
      ;; average cpu load is under the threshold before continuing

      ;; jobtools maxload is useful for where the full Megatest run is done on one machine
      (if maxload ;; only gate if maxload is specified, NOTE: maxload is normalized, i.e. load=1 means all cpus fully utilized
          (common:wait-for-normalized-load maxload "Waiting for load to drop before starting more tests" #f))

      ;; jobtools maxhomehostload is intended to prevent overloading on the homehost which can cause database corruption issues
      (if maxhomehostload
          (common:wait-for-homehost-load maxhomehostload (conc "Waiting for homehost load to drop below normalized value of " maxhomehostload)))
      
      (run:test run-id run-info keyvals runname test-record flags #f test-registry all-tests-registry)
      (runs:incremental-print-results run-id)
      (hash-table-set! test-registry (db:test-make-full-name test-name item-path) 'running)
      (runs:shrink-can-run-more-tests-count runsdat)  ;; DELAY TWEAKER (still needed?)

	     (conc (db:test-get-testname t) ":" (db:test-get-state t) "/" (db:test-get-status t))))
       lst))

;; parent-test is there as a placeholder for when parent-tests can be run as a setup step
;;
(define (run:test run-id run-info keyvals runname test-record flags parent-test test-registry all-tests-registry)
  ;; All these vars might be referenced by the testconfig file reader
  ;; 
  ;; NEED to reprocess testconfig here, ensuring that item variables are available.
  ;; This is for Tal's issue with item-specific env vars not being set for use in skip.
  ;; HSD https://hsdes.intel.com/appstore/icf/index.html#/article?articleId=1408763273
  ;;
  (let* ((test-name    (tests:testqueue-get-testname   test-record))
	 (test-waitons (tests:testqueue-get-waitons    test-record))

	 (itemdat      (tests:testqueue-get-itemdat    test-record))





	 (item-path     "")
	 (db           #f)
	 (full-test-name #f))

    ;; setting itemdat to a list if it is #f
    (if (not itemdat)(set! itemdat '()))
    (set! item-path (item-list->path itemdat))
    (set! full-test-name (db:test-make-full-name test-name item-path))
    (runs:set-megatest-env-vars run-id inrunname: runname testname: test-name itempath: item-path) ;; these may be needed by the launching process

   (let* ((test-conf ;; re-instate the tests:get-testconfig once the kinks are worked out. FIXME!!!
	    ;; (tests:get-testconfig test-name item-path all-tests-registry #t force-create: #t))
	    (tests:testqueue-get-testconfig test-record )) 
	   (test-path    (hash-table-ref all-tests-registry test-name)) ;; (conc *toppath* "/tests/" test-name)) ;; could use tests:get-testconfig here ...

	   (force        (hash-table-ref/default flags "-force" #f))
	   (rerun        (hash-table-ref/default flags "-rerun" #f))
	   (keepgoing    (hash-table-ref/default flags "-keepgoing" #f))
	   (incomplete-timeout (string->number (or (configf:lookup *configdat* "setup" "incomplete-timeout") "x")))
	   )

    (debug:print-info 4 *default-log-port*
		      "\nTESTNAME: " full-test-name 
		      "\n   test-config: " (hash-table->alist test-conf)
		      "\n   itemdat: " itemdat
		      )
    (debug:print 2 *default-log-port* "Attempting to launch test " full-test-name)
    ;; (setenv "MT_TEST_NAME" test-name) ;; 
    ;; (setenv "MT_ITEMPATH"  item-path)
    ;; (setenv "MT_RUNNAME"   runname)

    (change-directory *toppath*)

    ;; Here is where the test_meta table is best updated
    ;; Yes, another use of a global for caching. Need a better way?
    ;;
    ;; There is now a single call to runs:update-all-test_meta and this 
    ;; per-test call is not needed. Given the delicacy of the move to 

		  ((and skip-check
			(configf:lookup test-conf "skip" "prevrunning"))
		   ;; run-ids = #f means *all* runs
		   (let ((running-tests (rmt:get-tests-for-runs-mindata #f full-test-name '("RUNNING" "REMOTEHOSTSTART" "LAUNCHED") '() #f)))
		     (if (not (null? running-tests)) ;; have to skip 
			 (set! skip-test "Skipping due to previous tests running"))))

 		    ;; split the string and OR of file-exists?
		    ((and skip-check
			  (configf:lookup test-conf "skip" "fileexists"))
		     (let* ((files    (string-split (configf:lookup test-conf "skip" "fileexists")))
			    (existing (filter common:file-exists? files)))
		       (if (not (null? existing)) ;; (common:file-exists? (configf:lookup test-conf "skip" "fileexists"))
			   (set! skip-test (conc "Skipping due to existance of file(s) " (string-intersperse existing ", ")))))) ;; (configf:lookup test-conf "skip" "fileexists")))))

 		    ((and skip-check
			  (configf:lookup test-conf "skip" "filenotexists"))
		      (let* ((files    (string-split (configf:lookup test-conf "skip" "filenotexists")))
			      (existing (filter common:file-exists? files)))
			      (if (null? existing) ;;  (common:file-exists? (configf:lookup test-conf "skip" "filenotexists")))
			        (set! skip-test (conc "Skipping due to non existance of files " (string-intersperse files ", ")))))) ;; (configf:lookup test-conf "skip" "filenotexists")))))        
		 
       ((and skip-check
			(configf:lookup test-conf "skip" "script"))
		   (if (= (system (configf:lookup test-conf "skip" "script")) 0)
                        (set! skip-test (conc "Skipping due to zero return value of script " (configf:lookup test-conf "skip" "script")))))
      
                  ((and skip-check
			(configf:lookup test-conf "skip" "rundelay"))
		   ;; run-ids = #f means *all* runs

	;;      (debug:print 2 *default-log-port* "NOTE: " test-name " is already running")))
	(else      
	 (debug:print-error 0 *default-log-port* "Failed to launch test " full-test-name ". Unrecognised state " (test:get-state testdat))
	 (case (string->symbol (test:get-state testdat)) 
	   ((COMPLETED INCOMPLETE)
	    (hash-table-set! test-registry (db:test-make-full-name test-name test-path) 'DONOTRUN))
	   (else
	    (hash-table-set! test-registry (db:test-make-full-name test-name test-path) 'DONOTRUN)))))))))

;;======================================================================
;; END OF NEW STUFF
;;======================================================================

(define (get-dir-up-n dir . params) 
  (let ((dparts  (string-split dir "/"))

;; Copyright 2006-2017, Matthew Welland.
;; 
;; This file is part of Megatest.
;; 
;;     Megatest is free software: you can redistribute it and/or modify
;;     it under the terms of the GNU General Public License as published by
;;     the Free Software Foundation, either version 3 of the License, or

	     (not (equal? curr-ip target-host)))
	(begin
	  (debug:print-info 0 *default-log-port* "Starting server on " target-host ", logfile is " logfile)
	  (setenv "TARGETHOST" target-host)))
      
    (setenv "TARGETHOST_LOGF" logfile)
    (thread-sleep! (/ (random 5000) 1000)) ;; add about a random (up to 5 seconds) initial delay. It seems pretty common that many running tests request a server at the same time
    ;; (common:wait-for-normalized-load load-limit " delaying server start due to load" target-host) ;; do not try starting servers on an already overloaded machine, just wait forever
    (common:wait-for-homehost-load load-limit (conc " delaying server start due to load on homehost. limit is " load-limit))
    (system (conc "nbfake " cmdln))
    (unsetenv "TARGETHOST_LOGF")
    (if (get-environment-variable "TARGETHOST")(unsetenv "TARGETHOST"))
    (thread-join! log-rotate)
    (pop-directory)))

;; given a path to a server log return: host port startseconds