454099 - moz2-win32-slave03 and 5 need a good slapping

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P1)

Description

[Phil Ringnalda (:philor)]

Whatever that is that causes the Windows build slaves to leave a stray Fx process around, thus causing "nsinstall: cannot copy src\build\intel\mozcrt19.dll to ..\..\dist\bin\mozcrt19.dll: The process cannot access the file because it is being used by another process." and "nsinstall: cannot copy .\nspr4.dll to e:\builds\moz2_slave\mozilla-central-win32-debug\build\obj-firefox\dist\bin\nspr4.dll: The process cannot access the file because it is being used by another process.", moz2-win32-slave03 has it, and needs things terminated, and clobbered, and disappeared.

Comment 1

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

On the destkop a debug prompt of moz2-win32-slave03:
  The exception Breakpoint
  A breakpoint has been reached.
  (ox80000003) occured in the application at 0x7c81a3e1

firefox.exe is sitting there, so it looks like it failed to exit after the PGO launch.

moz2-win32-slave05 had the same thing with firefox.exe, but no prompt.

Both are fixed up, and I've forced a build for 
  WINNT 5.2 mozilla-central build %	
  WINNT 5.2 mozilla-central leak test build %

Comment 2

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

The leak-test build on moz2-win32-slave05 first went boom in the build at
2008/09/06 08:44:01. It was doing an alive test (cmd.exe /c python leaktest.py
-- -register):
   => mAllocCount:           6548
   => mReallocCount:           36
   => mFreeCount:            6531  --  LEAKED 17 !!!
   => mShareCount:           4579
   => mAdoptCount:            593
   => mAdoptFreeCount:        593
  ERROR: The process "10432" not found.

  command timed out: 1200 seconds without output
  SIGKILL failed to kill process
It should be proceeding to 
  SUCCESS: The process with PID <pid> has been terminated.
directly after th mAdoptFreeCount line.
---------------------------------------------------------

moz2-win32-slave03 leak test started having a problem during the 2008/09/07
08:12:59 build, doing a leak run (cmd.exe /c python leaktest.py --
--trace-malloc malloc.log --shutdown-leaks=sdleak.log):
  --DOMWINDOW == 9 (0BFCDB7C) [serial = 6] [outer = 00000000] [url =
http://localhost:8888/bloatcycle.html]
  --DOMWINDOW == 8 (0DC292C4) [serial = 10] [outer = 00000000] [url =
http://www.mozilla.org/newlayout/samples/test8.html]
  --DOMWINDOW == 7 (0DB2DBBC) [serial = 14] [outer = 00000000] [url =
http://www.mozilla.org/newlayout/samples/test8.html]
  --DOMWINDOW == 6 (0D4F680C) [serial = 9] [outer = 00000000] [url =
http://localhost:8888/bloatcycle.html]
  --WEBSHELL 0ABB6008 == 1
  WARNING: NS_ENSURE_SUCCESS(rv, rv) failed with result 0x80004005: file
e:/builds/moz2_slave/mozilla-central-win32-debug/build/security/manager/ssl/src/nsNSSIOLayer.cpp,
line 717

  command timed out: 3600 seconds without output
  SIGKILL failed to kill process
---------------------------------------------------------

The mozilla-central build on slave05 had issues from the 2008/09/07 05:47:47
build. Here the launch of Firefox for PGO failed to exit:
  OBJDIR=obj-firefox python obj-firefox/_profile/pgo/profileserver.py
  SSL tunnel pid: 11036
  Application pid: 13180
  Shutting down...
  failed to bind socket

  command timed out: 3600 seconds without output
  SIGKILL failed to kill process
The same issue started on on moz2-win32-slave03 at 2008/09/07 10:26:03, only it
doesn't have the Shutting down & failed to bind messages.

Something seems to be wrong in windows land.