Megatest

These are often the root cause of test problems on shared machines.

Remove runs on test_mt_vars and item 5 will not be deleted

Add locking (due to RUNNING etc.) on remove so that can safely remove symlinks

[ ] megatest -generate-test [ ] generate test issue - scripts vs. a command line. Look for |, space, quotes, / as signature of # Add additional steps here. Format is "stepname script" [ezsteps] getrepo fsl repo get `echo $TECH|sed 's/^p12/fdk/'` oalibs co_rel (cd oalibs;fsl co $RELEASE) convert convertall.sh Error: (open-output-file) cannot open file - No such file or directory: "../tests/csv_gen/fsl repo get `echo $TECH|sed 's/^p12/fdk/'` oalibs.sh"

Call history:

g210219 genexample.scm:240: print genexample.scm:241: print genexample.scm:242: get-environment-variable genexample.scm:242: print genexample.scm:243: get-environment-variable genexample.scm:243: print genexample.scm:244: print genexample.scm:245: print genexample.scm:246: print g245251 genexample.scm:251: conc genexample.scm:251: with-output-to-file genexample.scm:253: print genexample.scm:254: conc genexample.scm:254: with-output-to-file

Qns on megatest flow / methodology: 
1.	Turn off option for "Press Any key to continue" while clean/re-running tests in the xterm
1.	If I click execute again (accidentally),  will it re-queue the tests again?
1.	How to get dashboard with only the runs I want to see (only for 73 or 71 eg)
1.	Collapse/uncollapse all tests - option on dashboard required
1.	Checks are sorted order on website dashboard Vs unix dashboard unsorted
* Option to clean and rerun in one step on dashboard

look at the eval of [system ] blocks and see if it can be done per test/item vs. per test.

../envqa/tests/dumpndiffcdf/main.csh:megatest -test-status : state COMPLETED : status CHECK -setlog cdfdiff_prim/index.html

"SELECT id,run_id,testname,state,status,event_time,host,cpuload,diskfree,uname,rundir,item_path,run_duration,final_logf,comment FROM tests WHERE run_id=? AND ((testname LIKE '%if %'' AND item_path LIKE '%')) ;"

        Call history:

        db.scm:564: string-intersperse
        tests.scm:100: db:patt->like
        db.scm:560: ##sys#get-keyword
        db.scm:564: g779
        substring-index
        db.scm:566: conc
        db.scm:564: string-intersperse
        tests.scm:101: conc
        tests.scm:104: reverse
        tests.scm:104: append
        tests.scm:104: string-intersperse
        db.scm:796: conc
        db.scm:792: conc
        db.scm:802: debug:print-info
        common_records.scm:58: debug:debug-mode
        db.scm:803: sqlite3#for-each-row <--

==================
Now you need an initial place to store your runs. These are called "disks" and you
can add more at any time. To get going provide a writeable directory name. 
Enter your test runs directory: #{getenv MT_TEST_RUN_AREA}/../runs
WARNING: you have specified a path #{getenv MT_TEST_RUN_AREA}/../runs that does not exist. Attempting to create it...


Error: (create-directory) cannot create directory - File exists: "#{getenv MT_TEST_RUN_AREA}/.."

	Call history:

	genexample.scm:49: read-line	  
	genexample.scm:51: directory?	  
	genexample.scm:57: conc	  
	genexample.scm:57: file-exists?	  
	genexample.scm:68: print	  
	genexample.scm:81: display	  
	genexample.scm:82: read-line	  
	genexample.scm:84: string-split	  
	genexample.scm:84: g25	  
	genexample.scm:84: g25	  
	genexample.scm:87: print	  
	genexample.scm:89: display	  
	genexample.scm:90: read-line	  
	genexample.scm:91: directory?	  
	genexample.scm:93: print	  
	genexample.scm:94: create-directory	  	<--

[evn-override]
# BEQASCRIPTS #{scheme (print *toppath*)}/scripts

Launching /tmp/megatest/tests/fdktestqa/simplelinks/a/bigrun2/a/bigrun/890

Error: (hash-table-ref) hash-table does not contain key
tests-info
#<hash-table (2)>

	Call history:

	dashboard.scm:1053: iup-base#attribute-set!	  
	dashboard.scm:1055: iup-base#attribute-set!	  
	dashboard.scm:1342: current-milliseconds	  
	dashboard.scm:1389: mutex-lock!	  
	dashboard.scm:1391: mutex-unlock!	  
	dashboard.scm:1381: mutex-lock!	  
	dashboard.scm:1385: mutex-unlock!	  
	dashboard.scm:1388: dashboard:run-update	  
	dashboard.scm:1308: file-modification-time	  
	dashboard.scm:1309: file-modification-time	  
	dashboard.scm:1310: current-seconds	  
	dashboard.scm:1311: dashboard:recalc	  
	dashboard.scm:1296: current-milliseconds	  
	dashboard.scm:1389: mutex-lock!	  
	dashboard.scm:1391: mutex-unlock!	  
	dashboard.scm:813: hash-table-ref	

 cat NBFAKE-2013WW42.7_08\:42\:51 

Warning (#<thread: monitor job>): in thread: bad argument type - not a pointer: #f

        Call history:

        db.scm:277: k537          
        db.scm:277: g541          
        tests.scm:729: sqlite3#execute          <--

Error: uncaught exception: #<condition: (exn type)>

        Call history:

        launch.scm:358: thread-join!            <--

Warning (#<thread: ezstep run single step gencurrgds&g...Warning (#<thread: ezstep run single step gencurrgds>): in thread: unbound variable: run-id

Call history:

fs-transport.scm:42: debug:print-info common_records.scm:60: debug:debug-mode fs-transport.scm:43: db:process-queue-item db.scm:1849: g2314 db.scm:1850: g2318 db.scm:1851: g2322 db.scm:1852: g2326 db.scm:1853: alist-ref db.scm:1855: debug:print-info common_records.scm:60: debug:debug-mode db.scm:1865: debug:print-info common_records.scm:60: debug:debug-mode db.scm:1866: server:reply server.scm:108: debug:print-info common_records.scm:60: debug:debug-mode ezsteps.scm:167: cdb:roll-up-pass-fail-counts <--

shell-init: error retrieving current directory: getcwd: cannot access parent directories: No such file or directory 4:58 PM

chlr12706> megatest -step showresults -setlog results.html :state end :status 0
shell-init: error retrieving current directory: getcwd: cannot access parent directories: No such file or directory

Error: (current-directory) cannot retrieve current directory

        Call history:

        megatest.scm:316: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:333: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:348: hash-table-keys         
        megatest.scm:346: lset-intersection       
        megatest.scm:352: setup-for-run   
        launch.scm:422: hash-table?       
        launch.scm:425: args:get-arg      
        margs.scm:17: hash-table-ref/default      
        launch.scm:427: get-environment-variable          
        launch.scm:424: find-and-read-config      
        configf.scm:253: ##sys#get-keyword        
        configf.scm:253: ##sys#get-keyword        
        configf.scm:253: ##sys#get-keyword        
        configf.scm:254: current-directory              <--
chlr12706> cd $PWD
chlr12706> megatest -step showresults -setlog results.html :state end :status 0

[0] http-transport:client-connect -> ("192.168.0.35" 8080 #<request> #<request>) 
pid:1559 INFO: (1) Running -runstep, first change to directory /mfs/matt/data/megatest-inmem-with-per-run/tests/fullrun/tmp/mt_links/ubuntu/nfs/none/w08.7.21.45_aa/singletest

Error: (string-split) bad argument type - not a string: #f

        Call history:

        margs.scm:17: hash-table-ref/default      
        megatest.scm:931: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:934: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:937: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:940: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:946: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:947: args:get-arg    
        margs.scm:17: hash-table-ref/default      
        megatest.scm:948: conc    
        megatest.scm:952: get-environment-variable        
        megatest.scm:952: string-split          <--
; tracing tasks:server-lock-slot

Obtained lock for megatest-rollup-pycell_xor.html

... stuff ...

Error: uncaught exception: #<condition: (join-timeout-exception)>

Need to be able to automatically clean and rerun waitons.

Could also separate where -clean is a switch that modifies -runtests to clean before running.

INFO: (1) Test exited in state=RUNNING, setting state/status based on exit code of #t and rollup-status of 127

Error: (for-each) bad argument type - not a list: #<unspecified>

        Call history:

        http-transport.scm:274: thread-terminate!         
        http-transport.scm:275: debug:print-info          
        common_records.scm:68: debug:debug-mode   
        http-transport.scm:225: k682      
        http-transport.scm:225: g686      
        rmt.scm:59: db:string->obj        
        db.scm:1749: regex#regexp         
        db.scm:1748: regex#string-substitute      
        db.scm:1747: base64#base64-decode         
        db.scm:1746: with-input-from-string       
        db.scm:1750: s11n#deserialize     
        tests.scm:331: with-output-to-port        
        tests.scm:334: conc             <--
Obtained lock for megatest-rollup-sqlitespeed.html

Optional - make configurable

I don't know how the order of tests is decided in the megatest dashboard, but it looks like now expanding a test brings it to the bottom. It didn't used to do that, and I find it a little confusing. If it was not specifically requested by another user, I would recommend leaving the order the same, and just expanding in place.

Runs with failed tests showing as running when a test launches a new megatest instance.

I was helping a new user get set up today and I think we hit a first-use behavior in the dashboard. The behavior we noticed is that tests with test items start as expanded, which is fine. I had him select the collapse button and the test in progress collapsed. But as each successive test was populated with test items, that new test was presented as expanded. It appears that if there is nothing to collapse in a test and there is no previous regression using the test, the collapse is not applied.

Tests with items are rolling up to UNKNOWN state rather than PASS/FAIL/WARN as applicable.

This is happening with any number of items - I've tried 1 and 15.

When using itemmap and mode itemwait, if the mapped item contains a '%' the itemwait is behaving like toplevel - it is not starting the item when its dependencies have been met.

Example,

TESTA (waiton TESTB, mode itemwait) ITEM1/fluff ITEM2/fluff TESTB repo/ITEM1 repo/ITEM2

If I use itemmap to convert ITEM1/fluff to %/ITEM1, TESTA/ITEM1/fluff does not start until all items in TESTB are complete.

Its the central run dir. Pls let me know if you need more info.

Warning (#<thread: Main loop>): in thread: (finalize!) database is locked: #<sqlite3:statement sql="(5�">

Call history:

db.scm:3181: thread-sleep! db.scm:3182: db:delay-if-busy db.scm:3165: ##sys#get-keyword db.scm:3166: configf:lookup configf.scm:288: hash-table? configf.scm:289: hash-table-ref/default db.scm:3169: db:dbdat-get-path db.scm:3170: db:dbdat-get-db db.scm:3171: conc db.scm:3172: call-with-current-continuation db.scm:3172: with-exception-handler db.scm:3172: ##sys#call-with-values db.scm:3178: file-exists? db.scm:3172: k3877 db.scm:3172: g3881 db.scm:2039: sqlite3#for-each-row <--

Error: uncaught exception: #<condition: (exn sqlite3)>

1. capability to select entire row on dashboard 2. Select all visible runs on the dashboard and delete option 3. xterm sourcing setup.csh or some pre-command executed when launched from TCP.

When I expand a test with many subtests (200), if I use the scrollbar to scroll to the bottom, some tests still do not show up. This is intermittent. Sometimes if I scroll to the top and then back down, it then shows all tests; other times I cannot see the additional tests at all.

Run data intermittently disappears from the megatest dashboard. It can be recovered by closing the dashboard and reopening it, but "Refresh" does nothing.

When a testconfig is missing a specification of things to run, eg in [ezsetps] section, the test passes.

II would expect the test to fail since its specification is incomplete.

IIn my case, the error was a typo in spelling "ezsteps", but the test passed whereas it should have failed since the underlying step was actually failing but was masked by this issue.

Warning (#<thread: Main loop>): in thread: call of non-procedure: #f

Call history:

dashboard.scm:1853: file-modification-time dashboard.scm:1853: file-modification-time dashboard.scm:1853: file-modification-time dashboard.scm:1853: file-modification-time dashboard.scm:1853: file-modification-time dashboard.scm:1846: k3511 dashboard.scm:1846: g3515 dashboard.scm:1858: file-exists? dashboard.scm:1859: file-modification-time dashboard.scm:1861: current-seconds dashboard.scm:1862: d:alldat-please-update dashboard.scm:1862: d:alldat-last-db-update dashboard.scm:1862: dashboard:recalc dashboard.scm:1863: d:alldat-curr-tab-num dashboard.scm:1871: d:alldat-curr-tab-num dashboard.scm:1889: dashboard:update-new-view-tab <--

Error: uncaught exception: #<condition: (exn type)>

[6] Exit 70 dashboard -rows 25

in my testconfig, I have a step which runs a script having the code "exit 6"

In the test control panel, the Steps table correctly reflects Status=6 for the step.

However, the overall status is "FAIL". I expected "SKIP".

When having multiple runs, when I launch a new run, I click on the new run name that appears on the left pane of the Run Summary tab. The pain pane still shows results for the run that was previously selected. If I reselect the prior run, then select the new run again, it updates properly.

In megatest 1.6104 -list-targets attempts to create a symlink, so cannot check targets in another user's area.

NEEDS A TEST?

The sort +testname and -testname in dashboard are not sorting testname alphabetically - this should take the itempath into account if it exists.

For example, plxcm7003.

Hang occurs when switching to dashboard "run" or "run summary" tab

In Run tab, when viewing test names on leftmost column, the test names look like they have blank spaces where one would expect underscores. This was not the case in 1.60/29

(marking severe - reported by two users)

In the run tab, the width of matrix columns for runs is fixed. If a field value for the target or the run name spans more than ~14 characters, it is cut off.

This could be worked around in v1.60/29 by stretching the entire window horizontally by dragging the right corner out. Since the matrix was bound to the right side, the column widths would scale out. In current version (1.63/05) the matrix is not bound to the left edge, so stretching has no effect.

Marking Severe; reported by end user.

 mtutil run -area ext-tests -target a/b/c/d

Error: (null-list?) bad argument type - not a list: #<hash-table (2)>

        Call history:

        configf.scm:50: hash-table-ref/default
        configf.scm:49: config:assoc-safe-add
        configf.scm:41: ##sys#get-keyword
        configf.scm:42: filter
        configf.scm:43: append
        configf.scm:48: hash-table-set!
        mtut.scm:843: configf:lookup
        configf.scm:416: hash-table?
        configf.scm:417: hash-table-ref/default
        mtut.scm:844: string-split
        mtut.scm:845: hash-table-copy
        mtut.scm:841: ##sys#call-with-values
        mtut.scm:852: command-line->pkt
        mtut.scm:392: current-seconds
        mtut.scm:397: current-user-name
        mtut.scm:410: filter            <--

Since default and misc and % expand to non-sensical values and are not valid targets for runs, we do not expect to see these as selectable targets in run control panel.

I would like to propose a new [setup] option. In addition to max_concurrent_jobs, I would like to see something like max_waiting_jobs that would allow users to limit how many jobs are in the waiting or launched state. This would allow users to run as many jobs as possible, while still holding back the scheduling load of a batch system.

SETUP:
0) setup megatest build from tag v1.6407) 
1) under megatest folder, create ext-tests.  Poplulate subdir from megatest_qa fossil using commit d3e9c96dc307d7c941ea973d524adb4b267cbd56 (currently tip at time of writing)
2) Using megatest_qa testsuite, perform two runs, say named "run1" and "run2".  run1 completes first.  All tests in both runs should resolve to COMPLETED/PASS.
3) In run1, manually set tconfdisks/tom/n to RUNNING/PASS
5) in run1, manually set tconfdisks/fred/d to RUNNING/FAIL
6) in run1, manually set tconfdisks/bob/p to COMPLETED/FAIL
7) Make sure "HIDE" mode is selected for filters.
8) select state filter "RUNNING"
9) select status filter "FAIL"

OBSERVE:
two columns
a. run2 having tconfdisks topeleven and items tom/n and fred/d items as "RUNNING"
b. run1 having tconfdisks toplevel and items tom/n and fred/d as "RUNNING"
c. run1 having tconfdisks item bob/p "FAIL".

ISSUE:
item a above is unexpected, since they are in fact COMPLETED/PASS in the database.
item b and c are expected and correct.

current behavior when client version does not match server version is for client to launch another server. net result is a darwinian struggle of multiple version variants of megatest server.

motivation - it is hard to debug when a server has gone wrong. this is a feature indented for analysts and developers.

This caused confusion in a customer who did not understand the run indeed was deleted despite persisting in left pane hier browser.

Desire here is to have a broader view of the run. Not many columns and rows can fit on a single screen. If text were removed, then a cell could be only a few pixels tall and wide, allowing view of much more tests and items. Color is sufficient to indicate status.

A toggle or cyclic pulldown could select "normal" vs "condensed" view.

In an active run, the run summary matrix view induces headaches in yours truly. The column widths vary wildly from one second to the next. Eg: on second, the column fits width of "REMOTEHOSTSTART" and the next fits "PASS". multiply this delta by # of tests.

I have a test suite with tons of tests. When I did a series of "Clean Tests", I was expecting the list of runs on the left hand side to update to reflect the removed tests. It doesn't refresh until you kill the dashboard and restart it.

In runconfig files, common mistakes like creating a dividing line with ******** or accidentally wrapping a line, eg

yvar /blah/blah -ot foo.4.ot to myvar /blah/blah -ot foo<newline> .4.0t

Result in errors in csh and warnings in bash like ./megatest.sh: line 75: export: `.4.ot=': not a valid identifier

When creating megatest.csh/.sh, lets filter and warn when a env var name is invalid, that is does not begin with an alphabetic

I think a quick fix would be to add another button to context menu called CleanRunExecute! that does exactly what the test control panel button does and leave ReRun alone; if you dont trust one, or want more speed; choose your own fortune :)

Observations follow:

when i select test->rerun from right click menu in run tab of dashboard, the test doesnt seem to get cleaned before rerun i just noticed that in my crf build.. rerunning didnt clear steps... just ran them again... which looks weird since steps were out of order in test control panel with some #unspecified states

(iup:menu-item (conc "Rerun " item-test-path) #:action (lambda (obj) (common:run-a-command (conc "megatest -set-state-status NOT_STARTED,n/a -run -target " target " -runname " runname " -testpatt " item-test-path " -preclean -clean-cache")))) No Title (iup:menu-item "Run" (iup:menu (iup:menu-item (conc "Rerun " testpatt) #:action (lambda (obj) ;; (print " run-id: " run-id " test-id: " test-id " target: " target " runname: " runname " test-name: " test-name " testpatt: " testpatt "item-path : " item-path) (common:run-a-command (conc "megatest -run -target " target " -runname " runname " -testpatt " testpatt " -preclean -clean-cache") ))) (iup:menu-item "Rerun Complete Run" #:action (lambda (obj) (common:run-a-command (conc "megatest -set-state-status NOT_STARTED,n/a -run -target " target " -runname " runname " -testpatt % " " -preclean -clean-cache")))) all have -preclean

when I do clean-run-execute, it is explicitly cleared before retstarting... that is what guided my expectation

it definitely doesn't kill them... watching the dasboard, I see a running item go from running->launched immediately... with c-r-e, I see it blanking out first

(clean-run-execute (lambda (x) (let ((cmd (conc "megatest -remove-runs -target " keystring " -runname " runname " -testpatt " (conc testname "/" (if (equal? item-path "") "%" item-path)) ";megatest -target " keystring " -runname " runname " -run -preclean -testpatt " (conc testname "/" (if (equal? item-path "") "%" item-path)) " -clean-cache" )))

cre does a remove-runs first... that is the difference i guess

I think there might be some hazard. It ishard to prove... but I can get a test that passes consistently to fail if i rerun while running . In this case, the running test is in the process of a fossil open the respawned one dies with current directory is not within an open checkout

I

mt.conf

[fields]
a 
b
c
[setup]
...
[tasks]
job1 waitons=job2,job3;\
 logrules=(expect:required ...);\
 runtimlim=1m 5s; \
 envvars=A=BCD,C=DEF,;\
 steps=runtest.sh

#-OR-
[waitons]
job1 job2:im1
[itemmaps]
im1 .*/a \1

etc.

If linktree is changed in megatest.config, lt link in MT_RUN_AREA_HOME should be updated to match.

All the target definitions have "runs" in the run tree browser.

Example problem: from runconfigs.config: [full] ... /.*/ THINGS_LIST [system #{getenv THING_ORACLE_EXE} listem] OKSTUFF hello

% megatest -showrunconfig THINGS_LIST expected_value OKSTUFF hello

% megatest -showrunconfig -target full OKSTUFF hello **THINGS_LIST does not appear**

../

If no testpatt spec, then use the TESTPATT from runconfigs.

When a user puts a !FOO variable into an include in the runconfig that matches everything, like /.*/, it appears to match default as well, and when that happens, those variables leak into the environment.

If the target has been removed from the runconfig, the run does not show up in dashboard, and cannot be removed using the dashboard.

For easier querying and editing, split the content of pkts in the db into columns.

We need a way to prevent the pkts.db from getting too large. We need a way to remove all pkts that are no longer needed.

For a given target: If the youngest run consists of state=completed, status=<good>, then all older runs in same target can be eliminated.

I am using the megatest -list-test-time command to extract the run time for some regressions under my workarea.

When using the command the run times are provided but it does not seem to filter them out based on run name which makes the data difficult to understand.

In the log file additional test-patts are listed although only one has been specified.

Lastly, the documentation for this command states to use "-run-patt" and -test-patt" as options whic may be outdated.

Command: megatest -list-test-time -target $TARGET -runname my-run -testpatt test1

I think the current behavior in [skip] is to AND together multiple rules. Requesting capability to OR together multiple rules.

I am OK if the default remains AND but have a megatest.config setting or a different section header like [skipany] to implement.

Seeing instances of batch jobs running megatest far exceeding the runtime limit that is in the config. A couple of possible causes: megatest kills the job but the batch system doesn't detect that it's dead megatest does not honor the limit the limit could be getting overriden somewhere else that we're not seeing

For tests that have a step which is 2 commands joined with an && it's causing problems. The workaround for now is to split the command into two steps.

This is a request for an enhancement in megatest to be able to configure a test as a pre-requisite for all the tests within a flow without the need to manually configure the test dependency.

If i run: megatest -list-runs % -target % -fields tests:test_name,item_path

I expect to see only run and test info, not step info

When specifying -fields, step info should only be displayed when steps or +steps is specified.

The desired behavior is consistent with specifying run info in the fields - test info is suppressed then unless tests are specified

# shows runs, tests, and steps

-fields runs:runname+tests:test_name+steps:step_name

# shows runs and tests, no steps:

-fields runs:runname+tests:test_name+item_path

For script automation around -list-runs, it would also be nice if the test output looked like:

Test: cv item_path: c/lpp

Rather than:

Test: cv(c/lpp)

If it needs to be handled with a different dumpmode, that is ok too

skip is intended to short circuit a test

expected behavior:

• test: stop execution, mark test COMPLETED,SKIP, this is a "passing" status
• step: stop test execution, mark test COMPLTED,SKIP, this is a "passing" status

need to document use cases, eg how to "skip/short-circuit" a step but keep going

Requesting a new status: COMPLETED/STOP

The expected behavior is that the test stops, and subsequent items do not run.

This is a "soft pass" status color-wise (should not be red/purple)

Suggestion to implement via a [stop] testconfig section similar to [skip]

Need to have a way to regenerate the test logpro and scripts if the testconfig changes in MT_RUN_AREA_HOME/tests/<test name> area.

Not sure if it is necessary to be able to regenerate the files individually for each step.

When a variable is defined in a target earlier in the runconfig file, with no value, like

MYENVVAR

it appears with the name "MYENVVAR=" in the environment of a test run from a later target in the runconfig. It does not appear when doing megatest -show-runconfig -target <target>

The workaround is to put "" for the value of the var.

Need a megatest "dead man's switch" feature.

When applied, this would mark all pending operations in the database as "canceled" or something along those lines. Any runners calling home would not see any action left to take and exit.

Potentially should take target/runname/testpatt as arguments to filter which items to "cancel"

Looking to be able to override the logpro output directory for ezsteps.

Assuming the general functionality is:

step | logpro step.logpro step.html

The ask is to be able to add a testconfig (maybe also runconfig?) setting to redirect the html to a different location:

step | logpro step.logpro <override path>/step.html

From Robert: There is a real issue with the time it takes megatest to process runconfigs.

Every #{shell ...} command is processed in every target from every included file even if the target is not the one being run, or a simple -list-targets is given on the command line.

To make matters worse, when running a regression, not only does the main megatest "mtest" call processes runconfigs, so does any server started by the regression run.

And if it takes a LONG time to process the runconfigs, the regression run will spend a lot of time waiting for the server to start. Every single call to megatest, whether its to start the regression, or start a server, or whatever, is subject to the time it takes to process the runconfig files.

The parsing of runconfigs should happen 1 time per session, and the evaluation of value should only happen 1 time and when its needed/used (there should not be a global evaluation of all parameters across all targets unless there is some kind of dump operation occurring).

While I appreciate that there was some original reason/intent for re-parsing the runconfigs the entire process is leading to a substantial performance issues in regression runs. Even if there is a need to re-evaluate a variable value, there should be a means of indicating so, like:

VAR #{eval ...}

Something needs to be done to improve performance of reading the runconfigs.

Store vmstat data to see when the host is running low on memory

vmstat 5 10

procs -----------memory---------- ---swap-- -----io---- -system-- -----cpu------
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
 2  0 107184 13671396 1967828 73863552    0    1     0    33    0    0 18  1 80  1  0
 2  1 107184 13671476 1967828 73863552    0    0     0    22 4346 5371 11  1 88  1  0
 1  0 107184 13669244 1967828 73863552    0    0     0    45 4991 5907 13  1 86  1  0
 2  0 107184 13670680 1967828 73863552    0    0     0   198 4341 5371 11  1 87  1  0
 1  0 107184 13669668 1967828 73863552    0    0     0    32 4910 5509 11  1 87  1  0
 1  0 107184 13670744 1967828 73863552    0    0     0    68 4513 6001 11  1 87  1  0
 4  0 107184 13669328 1967828 73863552    0    0     0    33 5002 5906 11  1 87  1  0
 2  0 107184 13670512 1967828 73863552    0    0     0    40 4495 5793 11  1 87  1  0
 2  0 107184 13671172 1967828 73863552    0    0     0    53 4665 5230 11  1 87  2  0
 2  1 107184 13671704 1967828 73863552    0    0     0    15 4337 4935 11  1 88  1  0

megatest -run-from test1/item1

cleans all downstream tests and starts run from specified test

o	Error loading runconfigs that uses #{getenv USER} definition
	Failing to locate targets defined in unloaded runconfigs
o	–show-config => dumps (STDOUT) contents of megatest.config post-processing
o	–show-runconfig [–target <target>] => same but for runconfigs
o	–load <script> => scheme script (exits after execution)
o	–target <target> :runname <runname> -testpatt <testpatt> -waive <comment> -state COMPLETED –status FAIL
o	Reset FAIL -> NOT_STARTED
o	–target <target> :runname <runname> -set-state-status NOT_STARTED,n/a :state COMPLETED :status FAIL –testpatt %
	make fail_reset/fail_rerun

matt added on 2013-03-07 00:07:35 UTC:
Exact line in runconfigs.config is:

./runconfigs.config:[include ./configs/#{getenv USER}.runconfigs.config]

main test still says UNKNOWN If test is WARN, top-level should also be WARN ok. So SKIP is working as expected wrt item, but parent level test status is not correct (UNKNOWN)

Ok. So the rollup is not correct Let me check that.

For parent test, I think the conditions should be: All item PASS, parent PASS All item PASS or SKIP, parent PASS one item WARN, rest PASS/SKIP, parent WARN one item FAIL, parent FAIL one item WAIVED, no fail, parent WAIVED precedence is: FAIL, WARN, CHECK, PASS

The following rules need to apply to megatest for status for parent tests with items

1 FAIL item  parent test FAIL no FAIL, 1 CHECK  parent test CHECK no FAIL, no CHECK, 1 WAIVED  parent test WAIVED no FAIL, no CHECK, no WAIVED, 1 WARN  parent test WARN no FAIL, no CHECK, no WAIVED, no WARN, 1 SKIP  parent test PASS all SKIP  parent test SKIP all PASS  parent test PASS

Precedence order is:

FAIL => CHECK => WAIVED => WARN => PASS => SKIP

1 corner case:

If all items SKIP, parent is SKIP. So if total number of ITEMS and total number if SKIP match, then parent is SKIP.

SKIP is a lower priority than PASS as it reflects a preemptive cut of a PASS condition (nothing was actually checked/done) whereas PASS implies actual work to QA.

matt added on 2013-04-24 23:21:44 UTC:

      The following rules need to apply to megatest for status for
      parent tests with items

      1 FAIL item                                      parent test FAIL
      no FAIL, 1 CHECK                                 parent test CHECK
      no FAIL, no CHECK, 1 WAIVED                      parent test WAIVED
      no FAIL, no CHECK, no WAIVED, 1 WARN             parent test WARN
      no FAIL, no CHECK, no WAIVED, no WARN, 1 SKIP    parent test PASS
      all SKIP                                         parent test SKIP
      all PASS                                         parent test PASS

      Precedence order is:

      FAIL => CHECK => WAIVED => WARN => PASS => SKIP

      1 corner case:

      If all items SKIP, parent is SKIP. So if total number of ITEMS and
      total number if SKIP match, then parent is SKIP.

      SKIP is a lower priority than PASS as it reflects a preemptive cut
      of a PASS condition (nothing was actually checked/done) whereas PASS
      implies actual work to QA.

==================
Next we need to get the variables and values you wish to iterate this test over (blank for none)
Enter the variable name: EGG_NAME
Enter the space separated list of values for EGG_NAME: matchable readline apropos base64 regex-literals format "regex-case" "test" coops trace csv dot-locking posix-utils posix-extras directory-utils hostinfo tcp rpc csv-xml fmt json md5    
Enter the variable name: 
Hint; use .sh extension on the script names and we'll create placeholder scripts.#<unspecified>#<unspecified>#<unspecified>#<unspecified>

matt added on 2013-07-21 22:32:16 UTC:
More funky wording

==================
Now to enter the one or more steps that make up your test, note; you can add more later
Enter the name for this step (blank to stop):

matt added on 2013-07-23 04:06:14 UTC:
Script name mapping is not handled right. foo -> foo.sh or ??

Executable bits not set on scripts

Please extend itemmap support to support multiple itemmaps.

If more than one itemmmap exists in the [requirements] block, apply them sequentially top-down:

Ex:

With item: a/b/c

[requirements] ... itemmap [^/]+/ itemmap /.* -happy

Would result in:

b-happy

With multiple waiton dependencies, the items are applied only to the last waiton dep.

Figured out the minimum requirement to create the sxml files required by gnumeric to open a refdb. This will allow import from csv or direct creation of .dat files

The jobgroup setting in testconfig is currently defined in [test_meta] and does not do any variable interpolation.

The enhancement request is to add capability to set jobgroup when a test item is created so that variables set in [items] or [itemstable] are available.

megatest -list-runs % -state COMPLETED -status FAIL,CHECK does not filter on state/status

Requesting a changed to how megatest outputs tests to run

From a test window in dashboard we can click the "Start Xterm" button to bring up an xterm in the test run directory.

Would like to see this xterm have the environment for the test run automatically set.

megatest -set-mvals setup::max_concurrent_jobs=20,server::timeout=0.1 -set-rvals p1271/6/r2.3/iftag::FDKSETUP=’ fdk71 cfdk71.6_r2.3 qa -ot iftag.ot’

-set-mvals is megatest.config overrides
-set-rvals is runconfigs.config overrides

Syntax is:

section1::VAR1=VAL1,section2::VAR2=VAL2

Limitations:

1.	No commas, ever.
2.	No equals, ever.
3.	No double colons, ever.
4.	Values get applied once, we have to decide when.

refdb set command does not replace existing lines in the .dat file when updating existing keys:

refdb set db row col data

[] [col] row data

...

refdb set db row col data [] [col] row data row data

creating a new refdb with "refdb set" does not create the required sxml files to be able to open the new refdb via "refdb edit"

Non-dashboard megatest commands are requiring DISPLAY to be set

Megatest is on occasion getting into the state/status COMPLETED/UNKNOWN

changing status from fail to pass works changing status from fail to waived brings up the waive dialog, but on "apply and close" click, the message "Error: in callback: unbound variable: rmt:test-set-state-status-by" appears on the console, and dialog does not close. status does not change.

Following from xterm popped up from deleting runname 164am: (accessed by right clicking on a test in this run on dashboard->run->clean complete run

INFO: (1) Recursively removing /nfs/pdx/disks/icf_env_disk001/bjbarcla/gwa/issues/mtdev/randy-slow/reproduce/qa/libanabeqa/links/p1275/2/PDK_r2.0/libana/164am/oalib_propbag_diff

Error: (directory) cannot open directory - No such file or directory: "/nfs/pdx/disks/icf_env_disk001/bjbarcla/gwa/issues/mtdev/randy-slow/reproduce/qa/libanabeqa/links/p1275/2/PDK _r2.0/libana/164am/oalib_propbag_diff"

Call history:

common_records.scm:224: debug:debug-mode common_records.scm:225: port? common_records.scm:225: with-output-to-port runs.scm:1903: file-exists? runs.scm:1905: debug:print-info common_records.scm:224: debug:debug-mode common_records.scm:225: port? common_records.scm:225: with-output-to-port runs.scm:1906: file-exists? runs.scm:1907: runs:safe-delete-test-dir runs.scm:1662: directory-utils#directory-fold directory-utils.scm:128: ##sys#get-keyword directory-utils.scm:130: type-checks#check-procedure directory-utils.scm:132: check-directory directory-utils.scm:79: directory? directory-utils.scm:132: directory <--

If a test has items, requesting update to the state/status of the top-level test to roll up.

Current behavior in megatest is to look in runconfig for a TESTPATT setting if a testpatt is not specified on the command line.

This request is to enhance the behavior to add a -testpattvar parameter that takes a runconfig setting name to use for the default testpatt.

The testpattvar should be an optional argument that defaults to TESTPATT.

Sample usage: megatest -run -runname foo -target foo/bar -testpattvar FOOPATT.

I have observed that if you are running dashboard on a non-homehost it will work until the database size reaches some threshold, at which point it will freeze. When restarting, it immediately freezes. I suspect the database size is the critical dimension -- syncing likely takes longer than the action loop in dashboard.

In the instance I see it, the megatest.db size is 5.4M with 4k tests and 11 runs.

Running dashboard on homehost, there is no issue at all and dashboard responds very quickly.

Here is the console log from dashboard when we see the freezing issue:

Checking for user runconfigs config ... WARNING lefkowit.runconfigs.config file ...doesn't exist do this if you care ### Add custom targets/overrides to global targets here to ./configs/lefkowit.runconfigs.config

Checking for required/common linktree path: /p/fdk/gwa/lefkowit/qa/libanaacatqa/links ... exists /p/fdk/gwa/lefkowit/qa/libanaacatqa/links Defined disks: ------------------------------------ exists /p/fdk/gwa/lefkowit/qa/libanaacatqa/runs ------------------------------------ Checking if megatest variables are already defined ... STARTING DASHBOARD ... WARNING: Current policy requires running dashboard on homehost: (10.38.45.46 . #f) INFO: Trying to start server (megatest -server 10.38.45.46 -m testsuite:libanaacatqa) ... INFO: (0) Starting server on 10.38.45.46, logfile is /nfs/pdx/disks/ch_icf_megatest_001/lefkowit/mtTesting/megatestqa/libanaacatqa/logs/server.log #====================================================================== # NBFAKE logging command to: /nfs/pdx/disks/ch_icf_megatest_001/lefkowit/mtTesting/megatestqa/libanaacatqa/logs/server.log # megatest -server 10.38.45.46 -m testsuite:libanaacatqa #====================================================================== WARNING: problem with directory /p/fdk/gwa/lefkowit/fossil/ext/acatqa_ext/trunk/tests, dropping it from tests path WARNING: problem with directory /nfs/pdx/disks/ch_icf_megatest_001/lefkowit/mtTesting/megatestqa/relqa/tests, dropping it from tests path WARNING: problem with directory /nfs/pdx/disks/ch_icf_megatest_001/lefkowit/mtTesting/megatestqa/libanaacatqa/tests, dropping it from tests path NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 1000 to 2000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 2000 to 4000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 4000 to 8000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 8000 to 16000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 16000 to 32000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 32000 to 64000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 64000 to 128000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 128000 to 256000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 256000 to 512000 ... NOTE: increasing poll interval from 1125899906842624000 to 2251799813685248000 NOTE: updates are taking a long time, 3.0s elapsed. NOTE: increasing poll interval from 2251799813685248000 to 4503599627370496000 NOTE: updates are taking a long time, 3.0s elapsed.

Call history:

db.scm:1972: call-with-current-continuation db.scm:1972: with-exception-handler db.scm:1972: ##sys#call-with-values db.scm:1972: k3436 db.scm:1972: g3440 dashboard.scm:795: current-seconds dashboard.scm:798: dboard:tabdat-allruns-by-id dashboard.scm:798: hash-table-set! dashboard.scm:803: debug:print common_records.scm:131: debug:debug-mode common_records.scm:132: with-output-to-port dashboard.scm:804: iup-base#attribute dashboard.scm:805: floor dashboard.scm:3551: k7615 dashboard.scm:3551: g7619 dashboard.scm:3551: print-call-chain <-- inexact number cannot be represented as an exact number Callback error in dashboard:runs-tab-updater Full condition info: ((exn (location inexact->exact) (call-chain (#(db.scm:1978: loop #f #f) #(db.scm:1978: loop #f #f) #(db.scm:1978: loop #f #f) #(db.scm:1972: call-with-current-continuation #f #f) #(db.scm:1972: with-exception-handler #f #f) #(db.scm: 1972: ##sys#call-with-values #f #f) #(db.scm:1972: k3436 #f #f) #(db.scm:1972: g3440 #f #f) #(dashboard.scm:795: current-seconds #f #f) #(dashboard.scm:798: dboard:tabdat-allruns-by-id #f #f) #(dashboard.scm:798: hash-table-set! #f # f) #(dashboard.scm:803: debug:print #f #f) #(common_records.scm:131: debug:debug-mode #f #f) #(common_records.scm:132: with-output-to-port #f #f) #(dashboard.scm:804: iup-base#attribute #f #f) #(dashboard.scm:805: floor #f #f))) (arg uments (9.00719925474099e+18)) (message inexact number cannot be represented as an exact number)) (type)) NOTE: updates are taking a long time, 3.0s elapsed. <stack dump repeats regularly, probably in the dashboard event loop>

Hi, 1) Graphical representation of data will be very useful to increase productive 2) megatest enhancement which when one runs on a functional area shows graphically checks, fails, pass waivers 3) Then a deep dive graphical representation 4) megatest enhancement to show diff graphically mathematically between two different runs or kits (target1 run1 , target2 run2)

3) Kit health a. Number of waived test-items to start with. Percentage of waived testitems will be a very basic metric of kit health b. We should have infrastructure (some code) which will present us with Kit Health from highest level to lowest level of details for example i. At highest level it can give just percentage of waived ii. Also it can give diff wrt last kit released and improvement since then in some metric it will be nice iii. At lower level if we click on the graphs on waived section it should go into that part in details 1) e.g. waived should show % wrt tests like icv_lvs icv_drc etc 2) If we click on icv_drc it should show all tsetcases having issues 3) Another type of click on icv_drc should also show summary of different failing sub flows 4) All of above should be quantified c. This all should be done all functional areas i. Ofcourse many/most factors are different for all areas ii. I think we should have a excel sheet where users can input various factors and then use it to build graphs and , quantify results iii. Things should be quantified over functional areas and then over the entire kit (maths) iv. I think we should have a idea what we are aiming not just in English but numbers and then see where we reach what we miss d. Numbers will I guess make more sense over time than English

If a test in the test list for a run has a typo like "items]" instead of "[items]" (missing bracket) it causes an env check to fail at the top level (precheck test) megatest.csh will contain "#setenv items]", and precheck will flag an illegal env var.

Requesting a new -dumpmode option for -list-runs, -dumpmode shell

Purpose is to provide parseable output to include in shell scripts.

Requesting a default logpro section in testconfig. For a step, if there is no <step>.logpro

That is a tough one. It seems it could be considered a bug in Megatest. I’m treating a toplevel test in the same way as I’m treating items and non-itemized tests. What I need to do is not apply the remove to the FAIL toplevel if there are any non-fail items. This will be a bit tough and I’d like to wait until 1.60 to fix it.

Triggers log output via nbfake to $MT_TEST_RUN_DIR. This behavior works great for most triggers, but for REMOVING/ triggers $MT_TEST_RUN_DIR is being deleted.

Requesting for REMOVING/ triggers to log output to $MT_RUN_AREA_HOME/logs