282397 - MozillaTrunk symbols missing for Linux builds since mid-January

Status: RESOLVED FIXED

(Core Graveyard :: Talkback Client, defect)

Description

[Jay Patel [:jay]]

Looks like we haven't been getting symbols uploaded for Linux seamonkey builds
since January.  Chase is looking into it.

Comment 1

[Jay Patel [:jay]]

nominating for blocking1.8b

Comment 2

[Hermann Schwab]

see Bug 207136 Talkback data no longer reports register values for Windows builds

Bug 207136 Comment #3 From Shiva Thirumazhusai  2003-05-27 09:10 PST 

It is a known problem. 

Comment 3

[Jay Patel [:jay]]

Herman:  I'm pretty sure that other bug is referring to the register info in the
detailed Talkback reports (which can now only be viewed by internal Mozilla
staff).  This bug is for a server related issue that is causing us to get
unresolved stack frames in the Talkback reports everyone is able to lookup at
http://talkback-public.mozilla.org.  We simply need to figure out why the symbol
info for Linux Seamonkey builds are not being uploaded to our Talkback incident
processing server (so we can get nice looking stacks with functions, filenames,
and line numbers).

Comment 4

[Asa Dotzler [:asa]]

yep. gotta have this.

Comment 5

[Andrew Schultz]

this worksforme with linux trunk 2005021613
http://talkback-public.mozilla.org/talkback/fastfind.jsp?search=2&type=iid&id=TB3748014H

Comment 6

[Mats Palmgren (inactive)]

But the stack does not make sense. For example,
'ComputeBackgroundAnchorPoint' does not have a call to 'RenderSide'.

Comment 7

[Jay Patel [:jay]]

Talkback stacks don't always make sense...but the symbols are now there for
Linux builds, so marking this fixed.

Comment 8

[Mats Palmgren (inactive)]

(In reply to comment #7)
> Talkback stacks don't always make sense...

I am aware of that. It is unusual that the entire stack is unuseable though,
in my experience there are usually something at the beginning that does make
sense, like main() for example. The 5-6 latest talkback data I have submitted
are all unuseable (the entire stack does not make sense), that seems unlikely...

Could you please verify that symbols are working on Linux by manually doing
a crash using a abort() or something that does not mess up the stack. Thanks.

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

You've tried a build in the last few days?

Comment 10

[Chase Phillips]

(In reply to comment #8)
> I am aware of that. It is unusual that the entire stack is unuseable though,
> in my experience there are usually something at the beginning that does make
> sense, like main() for example. The 5-6 latest talkback data I have submitted
> are all unuseable (the entire stack does not make sense), that seems unlikely...

I think the GTK1 and GTK2 SeaMonkey builds are occasionally smashing each
other's symbols.  This happens if they end up with the same build ID (since
they're both Linux with the same product and version).  We'll probably need to
file a new bug for that if it turns out to be the case.

Comment 11

[Mats Palmgren (inactive)]

(In reply to comment #10)
> I think the GTK1 and GTK2 SeaMonkey builds are occasionally smashing each
> other's symbols.  This happens if they end up with the same build ID

Yes, that is the case - filed bug 282863.