Megatest

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

(use srfi-1 data-structures posix regex-case (prefix base64 base64:)
     format dot-locking csv-xml z3 udp ;; sql-de-lite
     hostinfo md5 message-digest typed-records directory-utils stack
     matchable regex posix (srfi 18) extras ;; tcp 
     (prefix nanomsg nmsg:)
     ;; (prefix nanomsg nmsg:)
     (prefix sqlite3 sqlite3:)
     pkts (prefix dbi dbi:)
     (prefix dbi dbi:)
     )

(declare (unit common))
(declare (uses commonmod))
(import (prefix commonmod cmod:))

(import commonmod)
(import pkts)

(include "common_records.scm")


;; (require-library margs)
;; (include "margs.scm")

	(max-allowed (string->number (or (configf:lookup *configdat* "setup" "max-logfiles") "300")))) ;; name -> age
    (if (not (directory-exists? "logs"))(create-directory "logs"))
    (directory-fold 
     (lambda (file rem)
       (handle-exceptions
	exn
	(begin
	  (debug:print-info 0 *default-log-port* "unable to rotate log " file ", probably handled by another process.")
	  (debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
	  (print-call-chain (current-error-port)))
	  (debug:print-info 2 *default-log-port* "unable to rotate log " file ", probably handled by another process, this is safe to ignore.")
	  (debug:print 2 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
	  ;; (print-call-chain (current-error-port)) ;; 
	  )
	(let* ((fullname  (conc "logs/" file))
	       (mod-time  (file-modification-time fullname))
	       (file-age  (- (current-seconds) mod-time)))
	  (hash-table-set! all-files file mod-time)
	  (if (or (and (string-match "^.*.log" file)
		       (> (file-size fullname) 200000))
		  (and (string-match "^server-.*.log" file)

    (if val val default)))

(define (assoc/default key lst . default)
  (let ((res (assoc key lst)))
    (if res (cadr res)(if (null? default) #f (car default)))))

(define (common:get-testsuite-name)
  (or (configf:lookup *configdat* "setup" "area-name") ;; megatest is a flexible tool, testsuite is too limiting a description.
      (configf:lookup *configdat* "setup" "testsuite" )
      (getenv "MT_TESTSUITE_NAME")
      (if (string? *toppath* )
  (cmod:get-testsuite-name *toppath* *configdat*))
          (pathname-file *toppath*)
          #f))) ;; (pathname-file (current-directory)))))

;; safe getting of toppath
(define (common:get-toppath areapath)
  (or *toppath*
      (if areapath
	  (begin
	    (set! *toppath* areapath)

;; use bash to expand a glob. Does NOT handle paths with spaces!
;;
(define (common:bash-glob instr)
  (string-split
   (with-input-from-pipe
       (conc "/bin/bash -c \"echo " instr "\"")
     read-line)))
       read-line)))

;;======================================================================
;; Some safety net stuff
;;======================================================================

;; return input if it is a list or return null
(define (common:list-or-null inlst #!key (ovrd #f)(message #f))
  (if (list? inlst)
      inlst
      (begin
	(if message (debug:print-error 0 *default-log-port* message))
	(or ovrd '()))))
  
;;======================================================================
;; T A R G E T S  ,   S T A T E ,   S T A T U S ,   
;;                    R U N N A M E    A N D   T E S T P A T T
;;======================================================================

;; (map print (map car (hash-table->alist (read-config "runconfigs.config" #f #t))))

		    (if (> curr-rownum rownum) curr-rownum rownum)
		    (if (> curr-colnum colnum) curr-colnum colnum)
		    ))))))

;; if it looks like a number -> convert it to a number, else return it
;;
(define (common:lazy-convert inval)
  (let* ((as-num (if (string? inval)(string->number inval) #f)))
    (or as-num inval)))
  (cmod:lazy-convert inval))

;; convert string a=1; b=2; c=a silly thing; d=
;; to '((a . 1)(b . 2)(c . "a silly thing")(d . ""))
;;
(define (common:val->alist val #!key (convert #f))
  (let ((val-list (string-split-fields ";\\s*" val #:infix)))
    (if val-list
	(map (lambda (x)
	       (let ((f (string-split-fields "\\s*=\\s*" x #:infix)))
		 (case (length f)
		   ((0) `(,#f))  ;; null string case
		   ((1) `(,(string->symbol (car f))))
		   ((2) `(,(string->symbol (car f)) . ,(let ((inval (cadr f)))
							 (if convert (common:lazy-convert inval) inval))))
  (cmod:val->alist val #!key (convert #f)))
		   (else f))))
	     val-list)
	'())))

;;======================================================================
;; S Y S T E M   S T U F F
;;======================================================================

;; lazy-safe get file mod time. on any error (file not existing etc.) return 0
;;

;; 			    (set! cpu-load newval))))))
;; 	      (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))
(define (common:get-cached-info key dtype #!key (age 10))
  (if *toppath*
      (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")
		     (debug:print-info 2 *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)))
      #f))
 
(define (common:write-cached-info key dtype dat)
  (if *toppath*
      (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)))))
      #f))
  

(define (common:raw-get-remote-host-load remote-host)
  (handle-exceptions
   exn
   #f ;; more specific handling of errors needed
   (with-input-from-pipe 
    (conc "ssh " remote-host " cat /proc/loadavg")
    (lambda ()(list (read)(read)(read))))))

;; get cpu load by reading from /proc/loadavg, return all three values
;;
(define (common:get-cpu-load remote-host)
  (handle-exceptions
   exn
   '(99 99 99)
   (let* ((actual-hostname (or remote-host (get-host-name) "localhost")))

					 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))))
	  (if (and (number? result)
		   (> result 0))
	  (if (> result 0)(common:write-cached-info actual-host "num-cpus" result))
	      (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 #!optional (rem-tries 5))
  (let ((num-cpus (common:get-num-cpus remote-host)))
    (if num-cpus

		      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
    ;; let's let the user know once in a long while that load checking is happening but not constantly report it
    (if (> (random 100) 75) ;; about 25% of the time
    (debug:print-info 1 *default-log-port* "Checking cpuload on " (or remote-host "localhost") ", maxload: " maxload
		      ", load: " first ", adjload: " adjload ", loadjmp: " loadjmp)
	(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)