bug#15863: Segmentation fault when traversing long lines

bug#15863: Segmentation fault when traversing long lines

[Aaron France]

[-- Attachment #1: Type: text/plain, Size: 695 bytes --]

Hi,

M-x version

GNU Emacs 24.3.1 (x86_64-unknown-linux-gnu, GTK+ Version 3.10.2)\n of 
2013-11-11 on archbox.

I've compiled this myself in order to retrieve debug symbols. Hence the 
date.

I seem to consistently be able to make emacs segfault in both the slime 
repl and the rgrep results window. Whenever the line is unusually long 
(how long? About 2-300 chars long?) and then go to the end of the line, 
it segfaults.

I've got the gdb backtrace, which is in this link: 
https://gist.github.com/AeroNotix/368ada4c84e7a52f461d

I've tried to trace it through but I am not familiar with this source 
and it's quite a length and involved file.

Any help would be greatly appreciated.

Aaron

[-- Attachment #2: Type: text/html, Size: 1168 bytes --]

bug#15863: Segmentation fault when traversing long lines

[Eli Zaretskii]

> Date: Mon, 11 Nov 2013 21:04:28 +0100
> From: Aaron France <aaron.l.france@gmail.com>
> 
> I seem to consistently be able to make emacs segfault in both the slime 
> repl and the rgrep results window. Whenever the line is unusually long 
> (how long? About 2-300 chars long?) and then go to the end of the line, 
> it segfaults.

If this is consistent, can you provide a reproducible test case,
starting with "emacs -Q"?

> I've got the gdb backtrace, which is in this link: 
> https://gist.github.com/AeroNotix/368ada4c84e7a52f461d

Thanks, but it's hard to do anything with just this backtrace.  Can
you at least see what was the immediate reason for the crash?  E.g.,
is 'glyph' a NULL pointer?

Also, how come you get truncation glyphs on a GUI frame? did you
disable the fringes or something?

Thanks.

bug#15863: Segmentation fault when traversing long lines

[Aaron France]

On 11/11/13 21:33, Eli Zaretskii wrote:
>> Date: Mon, 11 Nov 2013 21:04:28 +0100
>> From: Aaron France <aaron.l.france@gmail.com>
>>
>> I seem to consistently be able to make emacs segfault in both the slime
>> repl and the rgrep results window. Whenever the line is unusually long
>> (how long? About 2-300 chars long?) and then go to the end of the line,
>> it segfaults.
> If this is consistent, can you provide a reproducible test case,
> starting with "emacs -Q"?

"emacs" -Q does not reproduce the build.
>> I've got the gdb backtrace, which is in this link:
>> https://gist.github.com/AeroNotix/368ada4c84e7a52f461d
> Thanks, but it's hard to do anything with just this backtrace.  Can
> you at least see what was the immediate reason for the crash?  E.g.,
> is 'glyph' a NULL pointer?
>
> Also, how come you get truncation glyphs on a GUI frame? did you
> disable the fringes or something?
>
> Thanks.
I'll dig some more

bug#15863: Segmentation fault when traversing long lines

[Eli Zaretskii]

> Date: Mon, 11 Nov 2013 21:46:05 +0100
> From: Aaron France <aaron.l.france@gmail.com>
> CC: 15863@debbugs.gnu.org
> 
> On 11/11/13 21:33, Eli Zaretskii wrote:
> >> Date: Mon, 11 Nov 2013 21:04:28 +0100
> >> From: Aaron France <aaron.l.france@gmail.com>
> >>
> >> I seem to consistently be able to make emacs segfault in both the slime
> >> repl and the rgrep results window. Whenever the line is unusually long
> >> (how long? About 2-300 chars long?) and then go to the end of the line,
> >> it segfaults.
> > If this is consistent, can you provide a reproducible test case,
> > starting with "emacs -Q"?
> 
> "emacs" -Q does not reproduce the build.

Then perhaps you could add your customizations one by one until you
get to reproduce the problem, and then show the minimal set of
customizations that is enough to trigger it.

> I'll dig some more

Thank you.

bug#15863: Segmentation fault when traversing long lines

[npostavs]

tags 15863 unreproducible
close 15863
quit

Aaron France <aaron.l.france@gmail.com> writes:

> On 11/11/13 21:33, Eli Zaretskii wrote:
>>> Date: Mon, 11 Nov 2013 21:04:28 +0100
>>> From: Aaron France <aaron.l.france@gmail.com>
>>>
>>> I seem to consistently be able to make emacs segfault in both the slime
>>> repl and the rgrep results window. Whenever the line is unusually long
>>> (how long? About 2-300 chars long?) and then go to the end of the line,
>>> it segfaults.
>> If this is consistent, can you provide a reproducible test case,
>> starting with "emacs -Q"?
>
> "emacs" -Q does not reproduce the build.
> I'll dig some more

Since there's been no update for a while, I'm closing as unreproducible.