Megatest

Diff
Login

Differences From Artifact [0044077627]:

To Artifact [7b605ded00]:


75
76
77
78
79
80
81
82

83
84
85
86
87
88
89




90
91
92
93
94
95
96
75
76
77
78
79
80
81

82
83
84
85




86
87
88
89
90
91
92
93
94
95
96







-
+



-
-
-
-
+
+
+
+







    ;;   (if (and prevstep (file-exists? prev-env))
    ;;       (set! script (conc script "source " prev-env))))
    
    ;; call the command using mt_ezstep
    ;; (set! script (conc "mt_ezstep " stepname " " (if prevstep prevstep "x") " " stepcmd))
    
    (debug:print 4 "script: " script)
    (rmt:teststep-set-status! run-id test-id stepname "start" "-" #f #f)
    (rmt:teststep-set-status! run-id test-id stepname "start" "-" #f #f area-dat)
    ;; now launch the actual process
    (call-with-environment-variables 
     (list (cons "PATH" (conc (get-environment-variable "PATH") ":.")))
     (lambda ()
       (let* ((cmd (conc stepcmd " > " stepname ".log 2>&1")) ;; >outfile 2>&1 
	      (pid (process-run cmd)))
	 (rmt:test-set-top-process-pid run-id test-id pid)
     (lambda () ;; (process-run "/bin/bash" "-c" "exec ls -l /tmp/foobar > /tmp/delme-more.log 2>&1")
       (let* ((cmd (conc "exec " stepcmd " > " stepname ".log 2>&1")) ;; >outfile 2>&1 
	      (pid (process-run "/bin/bash" (list "-c" cmd))))
	 (rmt:test-set-top-process-pid run-id test-id pid area-dat)
	 (let processloop ((i 0))
	   (let-values (((pid-val exit-status exit-code)(process-wait pid #t)))
		       (mutex-lock! m)
		       (vector-set! exit-info 0 pid)
		       (vector-set! exit-info 1 exit-status)
		       (vector-set! exit-info 2 exit-code)
		       (mutex-unlock! m)
114
115
116
117
118
119
120
121

122
123

124
125
126
127
128
129
130
114
115
116
117
118
119
120

121
122

123
124
125
126
127
128
129
130







-
+

-
+







			    (begin
			      (thread-sleep! 2)
			      (processloop (+ i 1)))))
	    (debug:print-info 0 "logpro for step " stepname " exited with code " (vector-ref exit-info 2)))))
    
    (let ((exinfo (vector-ref exit-info 2))
	  (logfna (if logpro-used (conc stepname ".html") "")))
      (rmt:teststep-set-status! run-id test-id stepname "end" exinfo #f logfna))
      (rmt:teststep-set-status! run-id test-id stepname "end" exinfo #f logfna area-dat))
    (if logpro-used
	(rmt:test-set-log! run-id test-id (conc stepname ".html")))
	(rmt:test-set-log! run-id test-id (conc stepname ".html") area-dat))
    ;; set the test final status
    (let* ((this-step-status (cond
			      ((and (eq? (vector-ref exit-info 2) 2) logpro-used) 'warn)
			      ((eq? (vector-ref exit-info 2) 0)                   'pass)
			      (else 'fail)))
	   (overall-status   (cond
			      ((eq? (vector-ref exit-info 3) 2) 'warn) ;; rollup-status
155
156
157
158
159
160
161
162

163
164
165
166
167
168
169
155
156
157
158
159
160
161

162
163
164
165
166
167
168
169







-
+







	 (tests:test-set-status! run-id test-id next-state "PASS" #f #f))
	(else ;; 'fail
	 (vector-set! exit-info 3 1) ;; force fail, this used to be next-state but that doesn't make sense. should always be "COMPLETED" 
	 (tests:test-set-status! run-id test-id "COMPLETED" "FAIL" (conc "Failed at step " stepname) #f)
	 )))
    logpro-used))

(define (launch:execute encoded-cmd)
(define (launch:execute encoded-cmd area-dat)
  (let* ((cmdinfo   (common:read-encoded-string encoded-cmd)))
    (setenv "MT_CMDINFO" encoded-cmd)
    (if (list? cmdinfo) ;; ((testpath /tmp/mrwellan/jazzmind/src/example_run/tests/sqlitespeed)
	;; (test-name sqlitespeed) (runscript runscript.rb) (db-host localhost) (run-id 1))
	(let* ((testpath  (assoc/default 'testpath  cmdinfo))  ;; testpath is the test spec area
	       (top-path  (assoc/default 'toppath   cmdinfo))
	       (work-area (assoc/default 'work-area cmdinfo))  ;; work-area is the test run area
198
199
200
201
202
203
204
205


206
207

208
209



210
211
212
213
214
215
216
217


218
219

220
221
222
223
224
225

226
227
228
229
230
231
232


233
234
235
236
237
238
239
198
199
200
201
202
203
204

205
206
207
208
209


210
211
212
213
214
215
216
217
218


219
220
221

222
223
224
225
226
227

228
229
230
231
232
233


234
235
236
237
238
239
240
241
242







-
+
+


+
-
-
+
+
+






-
-
+
+

-
+





-
+





-
-
+
+







                                              runscript))))) ;; assume it is on the path
	       ;; (rollup-status 0)
	       )
	  (change-directory top-path)

	  ;; (set-signal-handler! signal/int (lambda ()
					    
	  ;; Do not run the test if it is REMOVING, RUNNING, KILLREQ or REMOTEHOSTSTART,
	  ;; WAS: Do not run the test if it is REMOVING, RUNNING, KILLREQ or REMOTEHOSTSTART,
	  ;; NOW: Do not run test test unless state is LAUNCHED
	  ;; Mark the test as REMOTEHOSTSTART *IMMEDIATELY*
	  ;;
	  ;; This is flawed. It should be a single transaction that tests for NOT_STARTED and updates to REMOTEHOSTSTART
	  (let ((test-info (rmt:get-testinfo-state-status run-id test-id)))
	    (if (not (member (db:test-get-state test-info) '("REMOVING" "REMOTEHOSTSTART" "RUNNING" "KILLREQ")))
	  (let ((test-info (rmt:get-testinfo-state-status run-id test-id area-dat)))
	  ;;
	    (if (equal? (db:test-get-state test-info) "LAUNCHED") ;; '("REMOVING" "REMOTEHOSTSTART" "RUNNING" "KILLREQ")))
		(tests:test-force-state-status! run-id test-id "REMOTEHOSTSTART" "n/a")
		(begin
		  (debug:print 0 "ERROR: test state is " (db:test-get-state test-info) ", cannot proceed")
		  (exit))))
	  
	  (debug:print 2 "Exectuing " test-name " (id: " test-id ") on " (get-host-name))
	  (set! keys       (rmt:get-keys))
	  ;; (runs:set-megatest-env-vars run-id inkeys: keys inkeyvals: keyvals) ;; these may be needed by the launching process
	  (set! keys       (rmt:get-keys area-dat))
	  ;; (runs:set-megatest-env-vars run-id area-dat inkeys: keys inkeyvals: keyvals) ;; these may be needed by the launching process
	  ;; one of these is defunct/redundant ...
	  (if (not (launch:setup-for-run force: #t))
	  (if (not (launch:setup-for-run area-dat force: #t))
	      (begin
		(debug:print 0 "Failed to setup, exiting") 
		;; (sqlite3:finalize! db)
		;; (sqlite3:finalize! tdb)
		(exit 1)))
	  (change-directory *toppath*) 
	  (change-directory toppath) 

	  ;; NOTE: Current order is to process runconfigs *before* setting the MT_ vars. This 
	  ;;       seems non-ideal but could well break stuff
	  ;;    BUG? BUG? BUG?

	  (let ((rconfig (full-runconfigs-read))) ;; (read-config (conc  *toppath* "/runconfigs.config") #f #t sections: (list "default" target))))
	    ;; (setup-env-defaults (conc *toppath* "/runconfigs.config") run-id (make-hash-table) keyvals target)
	  (let ((rconfig (full-runconfigs-read))) ;; (read-config (conc  toppath "/runconfigs.config") #f #t sections: (list "default" target))))
	    ;; (setup-env-defaults (conc toppath "/runconfigs.config") run-id (make-hash-table) keyvals target)
	    ;; (set-run-config-vars run-id keyvals target) ;; (db:get-target db run-id))
	    ;; Now have runconfigs data loaded, set environment vars
	    (for-each (lambda (section)
			(for-each (lambda (varval)
				    (let ((var (car varval))
					  (val (cadr varval)))
				      (if (and (string? var)(string? val))
270
271
272
273
274
275
276
277

278
279
280
281
282
283
284
285
286
287
288

289
290
291
292
293
294
295
273
274
275
276
277
278
279

280
281
282
283
284
285
286
287
288
289
290

291
292
293
294
295
296
297
298







-
+










-
+







	      (list  "MT_TEST_RUN_DIR" work-area)
	      (list  "MT_TEST_NAME" test-name)
	      (list  "MT_ITEM_INFO" (conc itemdat))
	      (list  "MT_ITEMPATH"  item-path)
	      (list  "MT_RUNNAME"   runname)
	      (list  "MT_MEGATEST"  megatest)
	      (list  "MT_TARGET"    target)
	      (list  "MT_LINKTREE"  (configf:lookup *configdat* "setup" "linktree"))
	      (list  "MT_LINKTREE"  (configf:lookup (megatest:area-configdat area-dat) "setup" "linktree"))
	      (list  "MT_TESTSUITENAME" (common:get-testsuite-name))))

	  (if mt-bindir-path (setenv "PATH" (conc (getenv "PATH") ":" mt-bindir-path)))
	  ;; (change-directory top-path)
	  ;; Can setup as client for server mode now
	  ;; (client:setup)

	  
	  ;; environment overrides are done *before* the remaining critical envars.
	  (alist->env-vars env-ovrd)
	  (runs:set-megatest-env-vars run-id inkeys: keys inkeyvals: keyvals)
	  (runs:set-megatest-env-vars run-id area-dat inkeys: keys inkeyvals: keyvals)
	  (set-item-env-vars itemdat)
	  (save-environment-as-files "megatest")
	  ;; open-run-close not needed for test-set-meta-info
	  ;; (tests:set-full-meta-info #f test-id run-id 0 work-area)
	  ;; (tests:set-full-meta-info test-id run-id 0 work-area)
	  (tests:set-full-meta-info #f test-id run-id 0 work-area 10)

317
318
319
320
321
322
323
324

325
326
327
328
329
330

331
332
333
334
335
336
337
320
321
322
323
324
325
326

327
328
329
330
331
332

333
334
335
336
337
338
339
340







-
+





-
+







				 ;; Since we should have a clean slate at this time there is no need to do 
				 ;; any of the other stuff that tests:test-set-status! does. Let's just 
				 ;; force RUNNING/n/a
				 

				 ;; (thread-sleep! 0.3)
				 (tests:test-force-state-status! run-id test-id "RUNNING" "n/a")
				 (rmt:roll-up-pass-fail-counts run-id test-name item-path "RUNNING")
				 (rmt:roll-up-pass-fail-counts run-id test-name item-path "RUNNING" area-dat)
				 ;; (thread-sleep! 0.3) ;; NFS slowness has caused grief here

				 ;; if there is a runscript do it first
				 (if fullrunscript
				     (let ((pid (process-run fullrunscript)))
				       (rmt:test-set-top-process-pid run-id test-id pid)
				       (rmt:test-set-top-process-pid run-id test-id pid area-dat)
				       (let loop ((i 0))
					 (let-values
					  (((pid-val exit-status exit-code) (process-wait pid #t)))
					  (mutex-lock! m)
					  (vector-set! exit-info 0 pid)
					  (vector-set! exit-info 1 exit-status)
					  (vector-set! exit-info 2 exit-code)
388
389
390
391
392
393
394
395

396
397
398
399
400
401
402
391
392
393
394
395
396
397

398
399
400
401
402
403
404
405







-
+







				       (if kill-job? 
					   (begin
					     (mutex-lock! m)
					     ;; NOTE: The pid can change as different steps are run. Do we need handshaking between this
					     ;;       section and the runit section? Or add a loop that tries three times with a 1/4 second
					     ;;       between tries?
					     (let* ((pid1 (vector-ref exit-info 0))
						    (pid2 (rmt:test-get-top-process-pid run-id test-id))
						    (pid2 (rmt:test-get-top-process-pid run-id test-id area-dat))
						    (pids (delete-duplicates (filter number? (list pid1 pid2)))))
					       (if (not (null? pids))
						   (begin
						     (for-each
						      (lambda (pid)
							(handle-exceptions
							 exn
443
444
445
446
447
448
449
450

451
452
453
454
455
456
457
446
447
448
449
450
451
452

453
454
455
456
457
458
459
460







-
+







	    (debug:print-info 0 "Megatest exectute of test " test-name ", item path " item-path " complete. Notifying the db ...")
	    (set! keep-going #f)
	    (thread-join! th1)
	    (thread-sleep! 1)       ;; give thread th1 a chance to be done TODO: Verify this is needed. At 0.1 I was getting fail to stop, increased to total of 1.1 sec.
	    (mutex-lock! m)
	    (let* ((item-path (item-list->path itemdat))
		   ;; only state and status needed - use lazy routine
		   (testinfo  (rmt:get-testinfo-state-status run-id test-id)))
		   (testinfo  (rmt:get-testinfo-state-status run-id test-id area-dat)))
	      ;; Am I completed?
	      (if (member (db:test-get-state testinfo) '("REMOTEHOSTSTART" "RUNNING")) ;; NOTE: It should *not* be REMOTEHOSTSTART but for reasons I don't yet understand it sometimes gets stuck in that state ;; (not (equal? (db:test-get-state testinfo) "COMPLETED"))
		  (let ((new-state  (if kill-job? "KILLED" "COMPLETED") ;; (if (eq? (vector-ref exit-info 2) 0) ;; exited with "good" status
				                                        ;; "COMPLETED"
							                ;; (db:test-get-state testinfo)))   ;; else preseve the state as set within the test
				    )
			(new-status (cond
480
481
482
483
484
485
486





















487
488

489
490
491
492

493

494
495
496
497
498
499
500
501
502
503
504

505
506
507
508


509
510

511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549







































550
551
552

553
554

555
556

557
558

559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581























582
583
584
585
586
587
588
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511

512
513
514
515
516
517

518











519




520
521


522







































523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563

564


565
566

567
568
569
570























571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600







+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

-
+




+
-
+
-
-
-
-
-
-
-
-
-
-
-
+
-
-
-
-
+
+
-
-
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+


-
+
-
-
+

-
+


+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+







	      (tests:summarize-test run-id test-id)) ;; don't force - just update if no
	    (mutex-unlock! m)
	    (debug:print 2 "Output from running " fullrunscript ", pid " (vector-ref exit-info 0) " in work area " 
			 work-area ":\n====\n exit code " (vector-ref exit-info 2) "\n" "====\n")
	    (if (not (vector-ref exit-info 1))
		(exit 4)))))))

(define (launch:read-cached-config)
  (if (get-environment-variable "MT_CMDINFO") ;; we are inside a test - do not reprocess configs
      (let ((alistconfig (conc (get-environment-variable "MT_LINKTREE") "/"
			       (get-environment-variable "MT_TARGET")   "/"
			       (get-environment-variable "MT_RUNNAME")  "/"
			       ".megatest.cfg")))
	(if (file-exists? alistconfig)
	    (list (configf:read-alist alistconfig)
		  (get-environment-variable "MT_RUN_AREA_HOME"))
	    #f))
      #f)) 

(define (launch:read-megatest-config toppath)
  (let ((runname (or (args:get-arg "-runname")(args:get-arg ":runname"))))
    (if runname (setenv "MT_RUNNAME" runname))
    (find-and-read-config 
     (if (args:get-arg "-config")(args:get-arg "-config") "megatest.config")
     environ-patt: "env-override"
     given-toppath: (get-environment-variable "MT_RUN_AREA_HOME")
     pathenvvar: "MT_RUN_AREA_HOME")))

;; set up the very basics needed for doing anything here.
(define (launch:setup-for-run #!key (force #f))
(define (launch:setup-for-run area-dat #!key (force #f))
  ;; would set values for KEYS in the environment here for better support of env-override but 
  ;; have chicken/egg scenario. need to read megatest.config then read it again. Going to 
  ;; pass on that idea for now
  ;; special case
  (let ((configdat (megatest:area-configdat area-dat)))
  (if (or force (not (hash-table? *configdat*)))  ;; no need to re-open on every call
    (if (or force (not (hash-table? configdat)))  ;; no need to re-open on every call
      (begin
	(set! *configinfo* (or (if (get-environment-variable "MT_CMDINFO") ;; we are inside a test - do not reprocess configs
				   (let ((alistconfig (conc (get-environment-variable "MT_LINKTREE") "/"
							    (get-environment-variable "MT_TARGET")   "/"
							    (get-environment-variable "MT_RUNNAME")  "/"
							    ".megatest.cfg")))
				     (if (file-exists? alistconfig)
					 (list (configf:read-alist alistconfig)
					       (get-environment-variable "MT_RUN_AREA_HOME"))
					 #f))
				   #f) ;; no config cached - give up
	(let* ((newconfiginfo (or (launch:read-cached-config) ;; no config cached - give up
			       (let ((runname (or (args:get-arg "-runname")(args:get-arg ":runname"))))
				 (if runname (setenv "MT_RUNNAME" runname))
				 (find-and-read-config 
				  (if (args:get-arg "-config")(args:get-arg "-config") "megatest.config")
				  (launch:read-megatest-config (megatest:area-path area-dat))))
	       (configdat     (car  newconfiginfo))
				  environ-patt: "env-override"
				  given-toppath: (get-environment-variable "MT_RUN_AREA_HOME")
	       (toppath       (cadr newconfiginfo)))
				  pathenvvar: "MT_RUN_AREA_HOME"))))
	(set! *configdat*  (if (car *configinfo*)(car *configinfo*) #f))
	(set! *toppath*    (if (car *configinfo*)(cadr *configinfo*) #f))
	(let* ((tmptransport (configf:lookup *configdat* "server" "transport"))
	       (transport    (if tmptransport (string->symbol tmptransport) 'http)))
	  (if (member transport '(http rpc nmsg))
	      (set! *transport-type* transport)
	      (begin
		(debug:print 0 "ERROR: Unrecognised transport " transport)
		(exit))))
	(let ((linktree (configf:lookup *configdat* "setup" "linktree"))) ;; link tree is critical
	  (if linktree
	      (if (not (file-exists? linktree))
		  (begin
		    (handle-exceptions
		     exn
		     (begin
		       (debug:print 0 "ERROR: Something went wrong when trying to create linktree dir at " linktree)
		       (debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
		       (exit 1))
		     (create-directory linktree #t))))
	      (begin
		(debug:print 0 "ERROR: linktree not defined in [setup] section of megatest.config")
		(exit 1)))
	  (if linktree
	      (let ((dbdir (conc linktree "/.db")))
		(handle-exceptions
		 exn
		 (begin
		   (debug:print 0 "ERROR: failed to create the " dbdir " area for your database files")
		   (debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn)))
		 (if (not (directory-exists? dbdir))(create-directory dbdir)))
		(setenv "MT_LINKTREE" linktree))
	      (begin
		(debug:print 0 "ERROR: linktree is required in your megatest.config [setup] section")
		(exit 1)))
	  (if (and *toppath*
		   (directory-exists? *toppath*))
	      (setenv "MT_RUN_AREA_HOME" *toppath*)
	  (megatest:area-configinfo-set! area-dat newconfiginfo)
	  (megatest:area-configdat-set!  area-dat configdat)
	  (megatest:area-path-set!       area-dat toppath)
	  (let* ((tmptransport (configf:lookup configdat "server" "transport"))
		 (transport    (if tmptransport (string->symbol tmptransport) 'http)))
	    (if (member transport '(http rpc nmsg))
		(megatest:area-transport-set! area-dat transport)
		(begin
		  (debug:print 0 "ERROR: Unrecognised transport " transport)
		  (exit))))
	  (let ((linktree (configf:lookup configdat "setup" "linktree"))) ;; link tree is critical
	    (if linktree
		(if (not (file-exists? linktree))
		    (begin
		      (handle-exceptions
		       exn
		       (begin
			 (debug:print 0 "ERROR: Something went wrong when trying to create linktree dir at " linktree)
			 (debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
			 (exit 1))
		       (create-directory linktree #t))))
		(begin
		  (debug:print 0 "ERROR: linktree not defined in [setup] section of megatest.config")
		  (exit 1)))
	    (if linktree
		(let ((dbdir (conc linktree "/.db")))
		  (handle-exceptions
		   exn
		   (begin
		     (debug:print 0 "ERROR: failed to create the " dbdir " area for your database files")
		     (debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn)))
		   (if (not (directory-exists? dbdir))(create-directory dbdir)))
		  (setenv "MT_LINKTREE" linktree))
		(begin
		  (debug:print 0 "ERROR: linktree is required in your megatest.config [setup] section")
		  (exit 1)))
	    (if (and toppath
		   (directory-exists? toppath))
	      (setenv "MT_RUN_AREA_HOME" toppath)
	      (begin
		(debug:print 0 "ERROR: failed to find the top path to your Megatest area.")
		(exit 1)))
		(exit 1))))
	  )))
  *toppath*)
	  toppath))))

(define (launch:cache-config)
(define (launch:cache-config area-dat)
  ;; if we have a linktree and -runtests and -target and the directory exists dump the config
  ;; to megatest-(current-seconds).cfg and symlink it to megatest.cfg
  (let ((configdat (megatest:area-configdat area-dat)))
  (if (and *configdat* 
	   (args:get-arg "-runtests"))
      (let* ((linktree (get-environment-variable "MT_LINKTREE"))
	     (target   (common:args-get-target))
	     (runname  (or (args:get-arg "-runname")
			   (args:get-arg ":runname")))
	     (fulldir  (conc linktree "/"
			     target "/"
			     runname)))
	(debug:print-info 0 "Have -runtests with target=" target ", runname=" runname ", fulldir=" fulldir)
	(if (file-exists? linktree) ;; can't proceed without linktree
	    (begin
	      (if (not (file-exists? fulldir))
		  (create-directory fulldir #t)) ;; need to protect with exception handler 
	      (if (and target
		       runname
		       (file-exists? fulldir))
		  (let ((tmpfile  (conc fulldir "/.megatest.cfg." (current-seconds)))
			(targfile (conc fulldir "/.megatest.cfg")))
		    (debug:print-info 0 "Caching megatest.config in " fulldir "/.megatest.cfg")
		    (configf:write-alist *configdat* tmpfile)
		    (system (conc "ln -sf " tmpfile " " targfile))
		    )))))))
    (if (and configdat 
	     (args:get-arg "-runtests"))
	(let* ((linktree (get-environment-variable "MT_LINKTREE"))
	       (target   (common:args-get-target))
	       (runname  (or (args:get-arg "-runname")
			     (args:get-arg ":runname")))
	       (fulldir  (conc linktree "/"
			       target "/"
			       runname)))
	  (debug:print-info 0 "Have -runtests with target=" target ", runname=" runname ", fulldir=" fulldir)
	  (if (file-exists? linktree) ;; can't proceed without linktree
	      (begin
		(if (not (file-exists? fulldir))
		    (create-directory fulldir #t)) ;; need to protect with exception handler 
		(if (and target
			 runname
			 (file-exists? fulldir))
		    (let ((tmpfile  (conc fulldir "/.megatest.cfg." (current-seconds)))
			  (targfile (conc fulldir "/.megatest.cfg")))
		      (debug:print-info 0 "Caching megatest.config in " fulldir "/.megatest.cfg")
		      (configf:write-alist configdat tmpfile)
		      (system (conc "ln -sf " tmpfile " " targfile))
		      ))))))))

(define (get-best-disk confdat)
  (let* ((disks    (hash-table-ref/default confdat "disks" #f))
	 (minspace (let ((m (configf:lookup confdat "setup" "minspace")))
		     (string->number (or m "10000")))))
    (if disks 
	(let ((res (common:get-disk-with-most-free-space disks minspace))) ;; min size of 1000, seems tad dumb
604
605
606
607
608
609
610
611
612



613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633


634
635
636
637
638
639
640
641
642

643
644
645
646
647
648
649
616
617
618
619
620
621
622


623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644


645
646
647
648
649
650
651
652
653
654

655
656
657
658
659
660
661
662







-
-
+
+
+



















-
-
+
+








-
+







;; 
;;  <linkdir> - <target> - <testname> [ - <itempath> ]
;; 
;; All log file links should be stored relative to the top of link path
;;  
;; <target> - <testname> [ - <itempath> ] 
;;
(define (create-work-area run-id run-info keyvals test-id test-src-path disk-path testname itemdat #!key (remtries 2))
  (let* ((item-path (if (string? itemdat) itemdat (item-list->path itemdat))) ;; if pass in string - just use it
(define (create-work-area run-id run-info keyvals test-id test-src-path disk-path testname itemdat area-dat #!key (remtries 2))
  (let* ((configdat (megatest:area-configdat area-dat))
	 (item-path (if (string? itemdat) itemdat (item-list->path itemdat))) ;; if pass in string - just use it
	 (runname   (if (string? run-info) ;; if we pass in a string as run-info use it as run-name.
			run-info
			(db:get-value-by-header (db:get-rows run-info)
						(db:get-header run-info)
						"runname")))
	 ;; convert back to db: from rdb: - this is always run at server end
	 (target   (string-intersperse (map cadr keyvals) "/"))

	 (not-iterated  (equal? "" item-path))

	 ;; all tests are found at <rundir>/test-base or <linkdir>/test-base
	 (testtop-base (conc target "/" runname "/" testname))
	 (test-base    (conc testtop-base (if not-iterated "" "/") item-path))

	 ;; nb// if itempath is not "" then it is prefixed with "/"
	 (toptest-path (conc disk-path "/" testtop-base))
	 (test-path    (conc disk-path "/" test-base))

	 ;; ensure this exists first as links to subtests must be created there
	 (linktree  (let ((rd (config-lookup *configdat* "setup" "linktree")))
		      (if rd rd (conc *toppath* "/runs"))))
	 (linktree  (let ((rd (config-lookup configdat "setup" "linktree")))
		      (if rd rd (conc (megatest:area-path area-dat) "/runs"))))

	 (lnkbase   (conc linktree "/" target "/" runname))
	 (lnkpath   (conc lnkbase "/" testname))
	 (lnkpathf  (conc lnkpath (if not-iterated "" "/") item-path))
	 (lnktarget (conc lnkpath "/" item-path)))

    ;; Update the rundir path in the test record for all, rundir=physical, shortdir=logical
    ;;                                                 rundir   shortdir
    (rmt:general-call 'test-set-rundir-shortdir run-id lnkpathf test-path testname item-path)
    (rmt:general-call 'test-set-rundir-shortdir run-id lnkpathf test-path testname item-path area-dat)

    (debug:print 2 "INFO:\n       lnkbase=" lnkbase "\n       lnkpath=" lnkpath "\n  toptest-path=" toptest-path "\n     test-path=" test-path)
    (if (not (file-exists? linktree))
	(begin
	  (debug:print 0 "WARNING: linktree did not exist! Creating it now at " linktree)
	  (create-directory linktree #t))) ;; (system (conc "mkdir -p " linktree))))
    ;; create the directory for the tests dir links, this is needed no matter what...
695
696
697
698
699
700
701
702

703
704
705
706
707
708
709
710
711
712
713
714

715
716
717
718
719
720
721
708
709
710
711
712
713
714

715
716
717
718
719
720
721
722
723
724
725
726

727
728
729
730
731
732
733
734







-
+











-
+







    ;; NB - This was not working right - some top tests are not getting the path set!!!
    ;;
    ;; Do the setting of this record after the paths are created so that the shortdir can 
    ;; be set to the real directory location. This is safer for future clean up if the link
    ;; tree is damaged or lost.
    ;; 
    (if (not (hash-table-ref/default *toptest-paths* testname #f))
	(let* ((testinfo       (rmt:get-test-info-by-id run-id test-id)) ;;  run-id testname item-path))
	(let* ((testinfo       (rmt:get-test-info-by-id run-id test-id area-dat)) ;;  run-id testname item-path))
	       (curr-test-path (if testinfo ;; (filedb:get-path *fdb*
							     ;; (db:get-path dbstruct
				   ;; (rmt:sdb-qry 'getstr 
				   (db:test-get-rundir testinfo) ;; ) ;; )
				   #f)))
	  (hash-table-set! *toptest-paths* testname curr-test-path)
	  ;; NB// Was this for the test or for the parent in an iterated test?
	  (rmt:general-call 'test-set-rundir-shortdir run-id lnkpath 
			    (if (file-exists? lnkpath)
				(resolve-pathname lnkpath)
				lnkpath)
			    testname "")
			    testname "" area-dat)
	  ;; (rmt:general-call 'test-set-rundir run-id lnkpath testname "") ;; toptest-path)
	  (if (or (not curr-test-path)
		  (not (directory-exists? toptest-path)))
	      (begin
		(debug:print-info 2 "Creating " toptest-path " and link " lnkpath)
		(handle-exceptions
		 exn
749
750
751
752
753
754
755
756

757
758
759
760
761
762
763
762
763
764
765
766
767
768

769
770
771
772
773
774
775
776







-
+







	   (if (not (file-exists? lnktarget)) (create-symbolic-link test-path lnktarget)))))

    (if (not (directory? test-path))
	(create-directory test-path #t)) ;; this is a hack, I don't know why out of the blue this path does not exist sometimes

    (if (and test-src-path (directory? test-path))
	(begin
	  (let* ((ovrcmd (let ((cmd (config-lookup *configdat* "setup" "testcopycmd")))
	  (let* ((ovrcmd (let ((cmd (config-lookup configdat "setup" "testcopycmd")))
			   (if cmd
			       ;; substitute the TEST_SRC_PATH and TEST_TARG_PATH
			       (string-substitute "TEST_TARG_PATH" test-path
						  (string-substitute "TEST_SRC_PATH" test-src-path cmd #t) #t)
			       #f)))
		 (cmd    (if ovrcmd 
			     ovrcmd
777
778
779
780
781
782
783
784
785




786
787
788

789
790
791
792
793
794

795
796
797
798
799
800

801
802
803
804
805
806


807
808

809
810
811
812
813
814
815
790
791
792
793
794
795
796


797
798
799
800
801
802

803
804
805
806
807
808

809
810
811
812
813
814

815
816
817
818
819


820
821
822

823
824
825
826
827
828
829
830







-
-
+
+
+
+


-
+





-
+





-
+




-
-
+
+

-
+







;; 1. look though disks list for disk with most space
;; 2. create run dir on disk, path name is meaningful
;; 3. create link from run dir to megatest runs area 
;; 4. remotely run the test on allocated host
;;    - could be ssh to host from hosts table (update regularly with load)
;;    - could be netbatch
;;      (launch-test db (cadr status) test-conf))
(define (launch-test test-id run-id run-info keyvals runname test-conf test-name test-path itemdat params)
  (change-directory *toppath*)
(define (launch-test test-id run-id run-info keyvals runname test-conf test-name test-path itemdat params area-dat)
  (let ((toppath   (megatest:area-path      area-dat))
	(configdat (megatest:area-configdat area-dat)))
  (change-directory toppath)
  (alist->env-vars ;; consolidate this code with the code in megatest.scm for "-execute"
   (list ;; (list "MT_TEST_RUN_DIR" work-area)
    (list "MT_RUN_AREA_HOME" *toppath*)
    (list "MT_RUN_AREA_HOME" toppath)
    (list "MT_TEST_NAME" test-name)
    ;; (list "MT_ITEM_INFO" (conc itemdat)) 
    (list "MT_RUNNAME"   runname)
    ;; (list "MT_TARGET"    mt_target)
    ))
  (let* ((useshell        (let ((ush (config-lookup *configdat* "jobtools"     "useshell")))
  (let* ((useshell        (let ((ush (config-lookup configdat "jobtools"     "useshell")))
			    (if ush 
				(if (equal? ush "no") ;; must use "no" to NOT use shell
				    #f
				    ush)
				#t)))     ;; default is yes
	 (launcher        (config-lookup *configdat* "jobtools"     "launcher"))
	 (launcher        (config-lookup configdat "jobtools"     "launcher"))
	 (runscript       (config-lookup test-conf   "setup"        "runscript"))
	 (ezsteps         (> (length (hash-table-ref/default test-conf "ezsteps" '())) 0)) ;; don't send all the steps, could be big
	 (diskspace       (config-lookup test-conf   "requirements" "diskspace"))
	 (memory          (config-lookup test-conf   "requirements" "memory"))
	 (hosts           (config-lookup *configdat* "jobtools"     "workhosts"))
	 (remote-megatest (config-lookup *configdat* "setup" "executable"))
	 (hosts           (config-lookup configdat "jobtools"     "workhosts"))
	 (remote-megatest (config-lookup configdat "setup" "executable"))
	 (run-time-limit  (or (configf:lookup  test-conf   "requirements" "runtimelim")
			      (configf:lookup  *configdat* "setup" "runtimelim")))
			      (configf:lookup  configdat "setup" "runtimelim")))
	 ;; FIXME SOMEDAY: not good how this is so obtuse, this hack is to 
	 ;;                allow running from dashboard. Extract the path
	 ;;                from the called megatest and convert dashboard
	 ;;             	  or dboard to megatest
	 (local-megatest  (let* ((lm  (car (argv)))
				 (dir (pathname-directory lm))
				 (exe (pathname-strip-directory lm)))
823
824
825
826
827
828
829
830

831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846

847
848
849
850
851


852
853
854
855
856
857
858
859
860
861
862
863
864
865
866

867
868

869
870
871
872
873
874
875
876
877
878
879
880

881
882
883
884
885
886



887






888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904

905
906
907
908
909
910
911
912
913
914
915
916
917

918
919
920
921
922
923
924
925

926
927
928
929
930
931
932

933
934
935
936
937
938
939
838
839
840
841
842
843
844

845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860

861
862
863
864


865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880

881
882

883
884
885
886
887
888
889
890
891
892
893
894

895
896
897
898
899
900
901
902
903
904

905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926

927
928
929
930
931
932
933
934
935
936
937
938
939

940
941
942
943
944
945
946
947

948
949
950
951
952
953
954

955
956
957
958
959
960
961
962







-
+















-
+



-
-
+
+














-
+

-
+











-
+






+
+
+
-
+
+
+
+
+
+
















-
+












-
+







-
+






-
+







	 (test-sig   (conc test-name ":" item-path)) ;; (item-list->path itemdat))) ;; test-path is the full path including the item-path
	 (work-area  #f)
	 (toptest-work-area #f) ;; for iterated tests the top test contains data relevant for all
	 (diskpath   #f)
	 (cmdparms   #f)
	 (fullcmd    #f) ;; (define a (with-output-to-string (lambda ()(write x))))
	 (mt-bindir-path #f)
	 (testinfo   (rmt:get-test-info-by-id run-id test-id))
	 (testinfo   (rmt:get-test-info-by-id run-id test-id area-dat))
	 (mt_target  (string-intersperse (map cadr keyvals) "/"))
	 (debug-param (append (if (args:get-arg "-debug")  (list "-debug" (args:get-arg "-debug")) '())
			      (if (args:get-arg "-logging")(list "-logging") '()))))
    (setenv "MT_ITEMPATH" item-path)
    (if hosts (set! hosts (string-split hosts)))
    ;; set the megatest to be called on the remote host
    (if (not remote-megatest)(set! remote-megatest local-megatest)) ;; "megatest"))
    (set! mt-bindir-path (pathname-directory remote-megatest))
    (if launcher (set! launcher (string-split launcher)))
    ;; set up the run work area for this test
    (if (and (args:get-arg "-preclean") ;; user has requested to preclean for this run
	     (not (member (db:test-get-rundir testinfo)(list "n/a" "/tmp/badname")))) ;; n/a is a placeholder and thus not a read dir
	(begin
	  (debug:print-info 0 "attempting to preclean directory " (db:test-get-rundir testinfo) " for test " test-name "/" item-path)
	  (runs:remove-test-directory testinfo 'remove-data-only))) ;; remove data only, do not perturb the record

    
    ;; prevent overlapping actions - set to LAUNCHED as early as possible
    ;;
    (tests:test-set-status! run-id test-id "LAUNCHED" "n/a" #f #f) ;; (if launch-results launch-results "FAILED"))
    (rmt:roll-up-pass-fail-counts run-id test-name item-path "LAUNCHED")
    (set! diskpath (get-best-disk *configdat*))
    (rmt:roll-up-pass-fail-counts run-id test-name item-path "LAUNCHED" area-dat)
    (set! diskpath (get-best-disk configdat))
    (if diskpath
	(let ((dat  (create-work-area run-id run-info keyvals test-id test-path diskpath test-name itemdat)))
	  (set! work-area (car dat))
	  (set! toptest-work-area (cadr dat))
	  (debug:print-info 2 "Using work area " work-area))
	(begin
	  (set! work-area (conc test-path "/tmp_run"))
	  (create-directory work-area #t)
	  (debug:print 0 "WARNING: No disk work area specified - running in the test directory under tmp_run")))
    (set! cmdparms (base64:base64-encode 
		    (z3:encode-buffer 
		     (with-output-to-string
		       (lambda () ;; (list 'hosts     hosts)
			 (write (list (list 'testpath  test-path)
				      (list 'transport (conc *transport-type*))
				      (list 'transport (conc (megatest:area-transport area-dat))) ;;
				      ;; (list 'serverinf *server-info*)
				      (list 'toppath   *toppath*)
				      (list 'toppath   toppath)
				      (list 'work-area work-area)
				      (list 'test-name test-name) 
				      (list 'runscript runscript) 
				      (list 'run-id    run-id   )
				      (list 'test-id   test-id  )
				      ;; (list 'item-path item-path )
				      (list 'itemdat   itemdat  )
				      (list 'megatest  remote-megatest)
				      (list 'ezsteps   ezsteps) 
				      (list 'target    mt_target)
				      (list 'runtlim   (if run-time-limit (common:hms-string->seconds run-time-limit) #f))
				      (list 'env-ovrd  (hash-table-ref/default *configdat* "env-override" '())) 
				      (list 'env-ovrd  (hash-table-ref/default configdat "env-override" '())) 
				      (list 'set-vars  (if params (hash-table-ref/default params "-setvars" #f)))
				      (list 'runname   runname)
				      (list 'mt-bindir-path mt-bindir-path))))))))

    ;; clean out step records from previous run if they exist
    ;; (rmt:delete-test-step-records run-id test-id)
    
    ;; Moving launch logs to MT_RUN_AREA_HOME/logs 
    ;;
    (change-directory work-area) ;; so that log files from the launch process don't clutter the test dir
    (let ((launchdir (configf:lookup *configdat* "setup" "launchdir"))) ;; (change-directory work-area) ;; so that log files from the launch process don't clutter the test dir
      (if (not launchdir) ;; default
	  (change-directory (conc *toppath* "/logs")) ;; can assume this exists
	  (case (string->symbol launchdir)
	    ((legacy)(change-directory work-area))
	    (else    (change-directory launchdir)))))
    (cond
     ((and launcher hosts) ;; must be using ssh hostname
      (set! fullcmd (append launcher (car hosts)(list remote-megatest test-sig "-execute" cmdparms) debug-param)))
     ;; (set! fullcmd (append launcher (car hosts)(list remote-megatest test-sig "-execute" cmdparms))))
     (launcher
      (set! fullcmd (append launcher (list remote-megatest test-sig "-execute" cmdparms) debug-param)))
     ;; (set! fullcmd (append launcher (list remote-megatest test-sig "-execute" cmdparms))))
     (else
      (if (not useshell)(debug:print 0 "WARNING: internal launching will not work well without \"useshell yes\" in your [jobtools] section"))
      (set! fullcmd (append (list remote-megatest test-sig "-execute" cmdparms) debug-param (list (if useshell "&" ""))))))
    ;; (set! fullcmd (list remote-megatest test-sig "-execute" cmdparms (if useshell "&" "")))))
    (if (args:get-arg "-xterm")(set! fullcmd (append fullcmd (list "-xterm"))))
    (debug:print 1 "Launching " work-area)
    ;; set pre-launch-env-vars before launching, keep the vars in prevvals and put the envionment back when done
    (debug:print 4 "fullcmd: " fullcmd)
    (let* ((commonprevvals (alist->env-vars
			    (hash-table-ref/default *configdat* "env-override" '())))
			    (hash-table-ref/default configdat "env-override" '())))
	   (testprevvals   (alist->env-vars
			    (hash-table-ref/default test-conf "pre-launch-env-overrides" '())))
	   (miscprevvals   (alist->env-vars ;; consolidate this code with the code in megatest.scm for "-execute"
			    (append (list (list "MT_TEST_RUN_DIR" work-area)
					  (list "MT_TEST_NAME" test-name)
					  (list "MT_ITEM_INFO" (conc itemdat)) 
					  (list "MT_RUNNAME"   runname)
					  (list "MT_TARGET"    mt_target)
					  (list "MT_ITEMPATH"  item-path)
					  )
				    itemdat)))
	   ;; Launchwait defaults to true, must override it to turn off wait
	   (launchwait     (if (equal? (configf:lookup *configdat* "setup" "launchwait") "no") #f #t))
	   (launchwait     (if (equal? (configf:lookup configdat "setup" "launchwait") "no") #f #t))
	   (launch-results (apply (if launchwait
				      cmd-run-with-stderr->list
				      process-run)
				  (if useshell
				      (let ((cmdstr (string-intersperse fullcmd " ")))
					(if launchwait
					    cmdstr
					    (conc cmdstr " >> mt_launch.log 2>&1")))
					    (conc cmdstr " >> " work-area "/mt_launch.log 2>&1")))
				      (car fullcmd))
				  (if useshell
				      '()
				      (cdr fullcmd)))))
      (if (not launchwait) ;; give the OS a little time to allow the process to start
	  (thread-sleep! 0.01))
      (with-output-to-file "mt_launch.log"
      (with-output-to-file (conc work-area "/mt_launch.log")
	(lambda ()
	  (if (list? launch-results)
	      (apply print launch-results)
	      (print "NOTE: launched \"" fullcmd "\"\n  but did not wait for it to proceed. Add the following to megatest.config \n[setup]\nlaunchwait yes\n  if you have problems with this"))
	  #:append))
      (debug:print 2 "Launching completed, updating db")
      (debug:print 2 "Launch results: " launch-results)
947
948
949
950
951
952
953
954
955




970
971
972
973
974
975
976


977
978
979
980







-
-
+
+
+
+
            ;; NB// Is this still needed? Should be safe to go back to "exit" now?
            (process-signal (current-process-id) signal/kill)
            ))
      (alist->env-vars miscprevvals)
      (alist->env-vars testprevvals)
      (alist->env-vars commonprevvals)
      launch-results))
  (change-directory *toppath*))

  (change-directory toppath))
  ;; added paren below after refactoring above routine. must have missed something?
  )