1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
|
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
|
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
(if (and test-src-path (> remtries 0))
(begin
(debug:print-error 0 *default-log-port* "Failed to create work area at " test-path " with link at " lnktarget ", remaining attempts " remtries)
;;
(create-work-area run-id run-info keyvals test-id test-src-path disk-path testname itemdat remtries: (- remtries 1)))
(list #f #f)))))
;;
(define (launch-test-standalone test-work-dir)
(when (not (directory-exists? test-work-dir))
(debug:print-error 0 *default-log-port* "Cannot launch. test-work-dir for lauched test does not exist, cannot proceed with launch: "test-work-dir)
(exit 1))
(change-directory test-work-dir)
(let* ((launch-dat-file (conc test-work-dir "/launch.dat")))
(if (not (common:file-exists? launch-dat-file))
;; error and exit
#f
(let* ((launch-info (with-input-from-file launch-dat-file read))
(run-id (alist-ref 'run-id launch-info))
(test-id (alist-ref 'test-id launch-info))
(work-area (alist-ref 'work-area launch-info))
(fullcmd (alist-ref 'fullcmd launch-info))
(launchwait (alist-ref 'launchwait launch-info))
(useshell (alist-ref 'useshell launch-info))
(launch-results-prev
(apply (if launchwait ;; BB: TODO: refactor this to examine return code of launcher, if nonzero, set state to launch failed.
process:cmd-run-with-stderr-and-exitcode->list
process-run)
(if useshell
(let ((cmdstr (string-intersperse fullcmd " ")))
(if launchwait
cmdstr
(conc cmdstr " >> mt_launch.log 2>&1 &")))
(car fullcmd))
(if useshell
'()
(cdr fullcmd))))
(success
(if launchwait (equal? 0 (cadr launch-results-prev)) #t))
(exit-code (if launchwait (cadr launch-results-prev) 0))
)
(if success
(tests:test-set-status! run-id test-id "LAUNCHED" "enqueued" #f #f)
(tests:test-set-status! run-id test-id "COMPLETED" "DEAD" "launcher failed; exited non-zero; check mt_launch.log" #f))
(with-output-to-file "mt_launch.log"
(lambda ()
(print "LAUNCHCMD: " (string-intersperse fullcmd " "))
(print "exit code => "exit-code)
(include "launch-test.scm")
#:append))
(debug:print 2 *default-log-port* "Launching completed, updating db")
(debug:print 2 *default-log-port* "Launcher exit code: " exit-code)
success))))
;; 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)
(mutex-lock! *launch-setup-mutex*) ;; setting variables and processing the testconfig is NOT thread-safe, reuse the launch-setup mutex
(let* ( ;; (lock-key (conc "test-" test-id))
;; (got-lock (let loop ((lock (rmt:no-sync-get-lock lock-key))
;; (expire-time (+ (current-seconds) 15))) ;; give up on getting the lock and steal it after 15 seconds
;; (if (car lock)
;; #t
;; (if (> (current-seconds) expire-time)
;; (begin
;; (debug:print-info 0 *default-log-port* "Timed out waiting for a lock to launch test " keyvals " " runname " " test-name " " test-path)
;; (rmt:no-sync-del! lock-key) ;; destroy the lock
;; (loop (rmt:no-sync-get-lock lock-key) expire-time)) ;;
;; (begin
;; (thread-sleep! 1)
;; (loop (rmt:no-sync-get-lock lock-key) expire-time))))))
(item-path (item-list->path itemdat))
(contour #f)) ;; NOT READY FOR THIS (args:get-arg "-contour")))
(let loop ((delta (- (current-seconds) *last-launch*))
(launch-delay (configf:lookup-number *configdat* "setup" "launch-delay" default: 1)))
(if (> launch-delay delta)
(begin
(if (common:low-noise-print 1200 "test launch delay") ;; every two hours or so remind the user about launch delay.
(debug:print-info 0 *default-log-port* "NOTE: test launches are delayed by " launch-delay " seconds. See megatest.config launch-delay setting to adjust.")) ;; launch of " test-name " for " (- launch-delay delta) " seconds"))
(thread-sleep! (- launch-delay delta))
(loop (- (current-seconds) *last-launch*) launch-delay))))
;;(change-directory *toppath*)
(alist->env-vars ;; consolidate this code with the code in megatest.scm for "-execute", *maybe* - the longer they are set the longer each launch takes (must be non-overlapping with the vars)
(append
(list
(list "MT_RUN_AREA_HOME" *toppath*)
(list "MT_TEST_NAME" test-name)
(list "MT_RUNNAME" runname)
(list "MT_ITEMPATH" item-path)
(list "MT_CONTOUR" contour)
)
itemdat))
(let* ((tregistry (tests:get-all)) ;; third param (below) is system-allowed
;; for tconfig, why do we allow fallback to test-conf?
(tconfig (or (tests:get-testconfig test-name item-path tregistry #t force-create: #t)
(begin
(debug:print 0 *default-log-port* "WARNING: falling back to pre-calculated testconfig. This is likely not desired.")
test-conf))) ;; force re-read now that all vars are set
(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
(runscript (config-lookup tconfig "setup" "runscript"))
(ezsteps (> (length (hash-table-ref/default tconfig "ezsteps" '())) 0)) ;; don't send all the steps, could be big
;; (diskspace (config-lookup tconfig "requirements" "diskspace"))
;; (memory (config-lookup tconfig "requirements" "memory"))
;; (hosts (config-lookup *configdat* "jobtools" "workhosts")) ;; I'm pretty sure this was never completed
(remote-megatest (config-lookup *configdat* "setup" "executable"))
(run-time-limit (or (configf:lookup tconfig "requirements" "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)))
(conc (if dir (conc dir "/") "")
(case (string->symbol exe)
((dboard) "../megatest")
((mtest) "../megatest")
((dashboard) "megatest")
(else exe)))))
(launcher (common:get-launcher *configdat* test-name item-path)) ;; (config-lookup *configdat* "jobtools" "launcher"))
(test-sig (conc (common:get-testsuite-name) ":" 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))
(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") '()))))
;; (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 *default-log-port* "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
;;
;; the following call handles waiver propogation. cannot yet condense into roll-up-pass-fail
(tests:test-set-status! run-id test-id "LAUNCHED" "n/a" #f #f) ;; (if launch-results launch-results "FAILED"))
(rmt:set-state-status-and-roll-up-items run-id test-name item-path #f "LAUNCHED" #f)
;; (pp (hash-table->alist tconfig))
(set! diskpath (get-best-disk *configdat* tconfig))
(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 *default-log-port* "Using work area " work-area))
(begin
(set! work-area (conc test-path "/tmp_run"))
(create-directory work-area #t)
(debug:print 0 *default-log-port* "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 'serverinf *server-info*)
(list 'homehost (let* ((hhdat (common:get-homehost)))
(if hhdat
(car hhdat)
#f)))
(list 'serverurl (if *runremote*
(remote-server-url *runremote*)
#f)) ;;
(list 'areaname (common:get-testsuite-name))
(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 'contour contour)
(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 '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)
;; if the dir does not exist we may have a itempath where individual variables are a path, launch anyway
(if (common:file-exists? work-area)
(change-directory work-area)) ;; so that log files from the launch process don't clutter the test dir
(cond
;; ((and launcher hosts) ;; must be using ssh hostname
;; (set! fullcmd (append launcher (car hosts)(list remote-megatest "-m" 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 "-m" test-sig "-execute" cmdparms) debug-param)))
;; (set! fullcmd (append launcher (list remote-megatest test-sig "-execute" cmdparms))))
(else
(if (not useshell)(debug:print 0 *default-log-port* "WARNING: internal launching will not work well without \"useshell yes\" in your [jobtools] section"))
(set! fullcmd (append (list remote-megatest "-m" 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 *default-log-port* "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 *default-log-port* "fullcmd: " fullcmd)
(set! *last-launch* (current-seconds)) ;; all that junk above takes time, set this as late as possible.
(let* ((commonprevvals (alist->env-vars
(hash-table-ref/default *configdat* "env-override" '())))
(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)))
(testprevvals (alist->env-vars
(hash-table-ref/default tconfig "pre-launch-env-overrides" '())))
;; Launchwait defaults to true, must override it to turn off wait
(launchwait (if (equal? (configf:lookup *configdat* "setup" "launchwait") "no") #f #t))
(launch-info (list
(cons 'run-id run-id)
(cons 'test-id test-id)
(cons 'work-area work-area)
(cons 'fullcmd fullcmd)
(cons 'launchwait launchwait)
(cons 'useshell useshell)))
(launch-dat-file (conc work-area "/launch.dat"))
(write-result (with-output-to-file launch-dat-file
(lambda () (pp launch-info))))
(launch-cmd (conc "megatest -start-dir "*toppath*" -internal-launch-test "work-area" &"))
)
(system launch-cmd)
(alist->env-vars miscprevvals)
(alist->env-vars testprevvals)
(alist->env-vars commonprevvals)
(change-directory *toppath*)
(mutex-unlock! *launch-setup-mutex*) ;; yes, really should mutex all the way to here. Need to put this entire process into a fork.
write-result))))
;; recover a test where the top controlling mtest may have died
;;
(define (launch:recover-test run-id test-id)
;; this function is called on the test run host via ssh
;;
;; 1. look at the process from pid
;; - is it owned by calling user
|