471295 - Leak test tinderbox turns red, but brief log summary remains empty

Status: RESOLVED DUPLICATE of bug 457976

(Webtools Graveyard :: Tinderbox, defect)

Description

[Serge Gautherie (:sgautherie)]

For example:
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1230420777.1230423016.7441.gz
WINNT 5.2 mozilla-central leak test build on 2008/12/27 15:32:57

NB: I don't even know what I should be looking for in the full log.

Comment 1

[Nick Thomas [:nthomas] (UTC+12)]

This is build is driven by buildbot so it's broken down into steps which follow the format:
  ======== BuildStep started ========
  <description of step>
  === Output ===
  <the command>
   in dir <working_dir> (timeout XX secs)
  watching logfiles {}
  argv: [ <for the command> ]
  environment:
  <long list of env vars here>
  <the actual output goes here>
  program finished with exit code 0
  === Output ended ===
  ======== BuildStep ended ========
To find problems you can search for non-zero exit codes.

In this case the problem is (abbreviated)
  === Output ===
  C:\WINDOWS\system32\cmd.exe /c hg pull -u
   in dir e:\builds\moz2-slave\mozilla-central-win32-debug\tools (timeout 1200 secs)

  pulling from http://hg.mozilla.org/build/tools
  waiting for lock on repository e:\builds\moz2-slave\mozilla-central-win32-debug\tools held by ''
  abort: repository e:\builds\moz2-slave\mozilla-central-win32-debug\tools: timed out waiting for lock held by 
  program finished with exit code -1

The tinderbox server is applying the "unix" error parser to this (see the very top of the full log), which is at
  http://mxr.mozilla.org/mozilla/source/webtools/tinderbox/ep_unix.pl
If you want to teach it new tricks then this bug should be moved over to Webtools::Tinderbox.

I can't see any rogue processes on moz2-win32-slave1, which was the slave used when the error above was generated. It's just successfully done hg operations in different dir on the same slave, but I could reproduce the error and just blew away the tools checkout.

Comment 2

[Serge Gautherie (:sgautherie)]

(In reply to comment #1)
> To find problems you can search for non-zero exit codes.

Thanks.

> If you want to teach it new tricks then this bug should be moved over to
> Webtools::Tinderbox.

Yes, that's what I want in this bug.

Comment 3

[cls]

Ugh.  First, this is a dupe.  Second, buildbot is building on windows but using the unix parser.  Third, buildbot is not using any of the standard lines to indicate build failure for the parser that it's using.

Comment 4

[Serge Gautherie (:sgautherie)]

(In reply to comment #3)

> First, this is a dupe.

Is it ?
This bug is about the generic "program finished with exit code ...";
not bug 457976 specific (Hg) "abort: ...".

> Second, buildbot is building on windows but using the unix parser.

Is this something wanted or to fix ?

> Third, buildbot is not using any of the standard lines to
> indicate build failure for the parser that it's using.

Is this the "program finished with exit code ..." issue !?
Would we need an ep_leaktest.pl ?

Comment 5

[Serge Gautherie (:sgautherie)]

(In reply to comment #4)
> not bug 457976 specific (Hg) "abort: ...".

s/Hg/Python/