346843 - enable nsSigHandlers trapping for x86_64

Status: RESOLVED FIXED

(Core :: XUL, defect)

Description

[Sylvain Pasche]

User-Agent:       Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.8.1b1) Gecko/20060801 BonEcho/2.0b1
Build Identifier: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.8.1b1) Gecko/20060801 BonEcho/2.0b1

patch coming

Reproducible: Always

Comment 1

[Sylvain Pasche]

Attachment: patch, v1

That is working fine for me. But I don't know it was not enabled for other more complex reasons.

Comment 2

[Sylvain Pasche]

s/know it was/know if it was/

Comment 3

[Sylvain Pasche]

Comment on attachment 231575 [details] [diff] [review]
patch, v1

I think I forgot to ask for super-review (I took bug 333538 as example). Can you commit that afterwards?

Comment 4

[Brian Ryner (not reading)]

Checked in, thanks.

Comment 5

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

It's probably severely broken in debug+opt builds on newer gcc's -- at least it was the last time I tested.

Comment 6

[Sylvain Pasche]

(In reply to comment #5)
> It's probably severely broken in debug+opt builds on newer gcc's -- at least it
> was the last time I tested.
> 

I have been testing this with gcc 4.0. Maybe I will try with gcc 4.1.

By "severly broken", you mean that the adresses reported by the stacktrace are not correct?
I can attach some output there if helpful to diagnose the issues.

Comment 7

[Sylvain Pasche]

Attachment: sample stack trace output on x86_64, gcc 4.1

Comment 8

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

I mean that not only are large numbers of stack frames missing, but it often crashed somewhere in the middle of the stack when EBP was used for something other than the frame pointer, especially when the stack went through system libraries which often ship without the frame pointer.

Comment 9

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

*** Bug 347287 has been marked as a duplicate of this bug. ***