677033 - Since 20110805 Nightly8.0a1, Crash Reporter always sends signature: xul.dll@xxxxx, does not send exact signature.

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P2)

Description

[Alice0775 White]

Build Identifier:
http://hg.mozilla.org/mozilla-central/rev/be090ee1747a
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:8.0a1) Gecko/20110806 Firefox/8.0a1 ID:20110806030740

Since 20110805 Nightly8.0a2, Crash Reporter always sends signature: xul.dll@xxxxx like a hourly build, does not send not exact signature.

Reproducible: Always

Steps to Reproduce:

1. Start browser
2. Crash browser
3.

Actual Results: 
  always sends signature: xul.dll@xxxxx

Expected Results: 
  Should send exact signature.

Comment 1

[Alice0775 White]

regression window:
Works:
http://hg.mozilla.org/mozilla-central/rev/be4b064f1159
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:8.0a1) Gecko/20110804 Firefox/8.0a1 ID:20110804030732
Fails:
http://hg.mozilla.org/mozilla-central/rev/5684f06138f3
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:8.0a1) Gecko/20110805 Firefox/8.0a1 ID:20110805030744
pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=be4b064f1159&tochange=5684f06138f3

Comment 2

[Josh Matthews [:jdm]]

Bug 674888 looks like the likely culprit.

Comment 3

[(not currently active) Ted Mielczarek]

That doesn't seem very likely at all. This regression is on Windows, and elfhack is for Linux...

Comment 4

[(not currently active) Ted Mielczarek]

There's definitely something broken in the symbol files.
Here's the 2011-08-04 nightly:
[tmielczarek@dm-symbolpush01 symbols_ffx]$ ls *8.0*WINNT*20110804*
firefox-8.0a1-WINNT-20110804030732-symbols.txt

$ grep xul firefox-8.0a1-WINNT-20110804030732-symbols.txt
xul.pdb/318E544510FB4AC18A9D1EA389E96EA02/xul.sym
xul.pdb/318E544510FB4AC18A9D1EA389E96EA02/xul.pd_

$ wc -l xul.pdb/318E544510FB4AC18A9D1EA389E96EA02/xul.sym
913346 xul.pdb/318E544510FB4AC18A9D1EA389E96EA02/xul.sym

$ grep FUNC xul.pdb/318E544510FB4AC18A9D1EA389E96EA02/xul.sym | wc -l
94942

$ grep STACK xul.pdb/318E544510FB4AC18A9D1EA389E96EA02/xul.sym | wc -l
49235

compare to the 2011-08-05 nightly:
$ ls *8.0*WINNT*20110805*
firefox-8.0a1-WINNT-20110805030744-symbols.txt

$ grep xul firefox-8.0a1-WINNT-20110805030744-symbols.txt
xul.pdb/2B6071C4E2AA4ADE9E1BDF997836E06F2/xul.sym
xul.pdb/2B6071C4E2AA4ADE9E1BDF997836E06F2/xul.pd_

$ wc -l xul.pdb/2B6071C4E2AA4ADE9E1BDF997836E06F2/xul.sym
947617 xul.pdb/2B6071C4E2AA4ADE9E1BDF997836E06F2/xul.sym

$ grep FUNC xul.pdb/2B6071C4E2AA4ADE9E1BDF997836E06F2/xul.sym | wc -l
103263

$ grep STACK xul.pdb/2B6071C4E2AA4ADE9E1BDF997836E06F2/xul.sym | wc -l
0

I don't know exactly what's going wrong though.

Comment 5

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

Looking at the buildbot log for the Aug 5th nightly (m-c, win32) [1] then the magic id for xul.pdb is A0E701116B654E4DAB312A879E1962532 rather than 2B6071..., and it turns out that the Win64 and Win32 builds are overwriting each others manifest files (symbols.txt). The Win64 sym files don't have any STACK lines (is that a bug?). 

Since we still have separate sym files for the two platforms the crash should still be resolved on Socorro. We've also been overwriting the manifests for many days.

[1] http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-central-win32/1312538864/mozilla-central-win32-nightly-build9.txt.gz

Comment 6

[Alice0775 White]

Today's Nightly8.0a1 seems to work properly again.

http://hg.mozilla.org/mozilla-central/rev/b422fd99fe0d
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:8.0a1) Gecko/20110807 Firefox/8.0a1 ID:20110807030727

Comment 7

[(not currently active) Ted Mielczarek]

Nick: oh, good call. We should fix that manifest overwriting though, because it is going to screw our ability to clean up symbols...

(The Win64 symbols not having STACK lines is bug 548035.)

Comment 8

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

Bug 677116 for the manifest naming fix.

Comment 9

[Armen [:armenzg]]

This did get fixed now that bug 677116 got fixed?

Comment 10

[Armen [:armenzg]]

Please reopen if it did not get fixed.