Re: [bug-libextractor] Buffer overflow detected

[Christian Grothoff]

Thanks for the update, forwarding the list to archive the result of the
investigation.

On 11/17/2015 01:14 PM, David Earl wrote:
> Hi,
> 
> I thought I'd better let you know that it does appear that this problem
> was probably due to a hardware fault, somewhat to my surprise. After
> this, there was a quiet period and then other strange events started
> happening including some kernel and interrupt faults which were getting
> more frequent, and which were really unlikely to be bugs. In the end my
> provider decided the problem probably was RAM related as you speculated,
> Christian. We transferred everything to new hardware yesterday. So,
> thank you for your help. I'm generally wary of attributing things like
> this to real hardware faults, but it does look like that's what it was
> on this occasion.
> 
> David
> 
> On Wed, 28 Oct 2015 at 23:12 Christian Grothoff <address@hidden
> <mailto:address@hidden>> wrote:
> 
>     On 10/28/2015 11:11 PM, David Earl wrote:
>     > So it looks like I've perturbed it by doing this in such a way
>     that it has
>     > hidden the problem.
> 
>     Hmm.  Just to be sure: have you checked that the RAM in your system is
>     OK? I've seen this where things go haywire at some point because some
>     bit in the RAM is bad, and then if the application is loaded in just the
>     right way, it goes to hell, but otherwise all seems fine.
> 
>     Alternatively, you can try to narrow things down by disabling (i.e.
>     removing) libextractor plugins (i.e. /usr/lib/libextractor/*.so). Remove
>     those files (or rather, probably, "move") one-by-one. That way we ought
>     to be able to tell which specific plugin causes trouble (I don't expect
>     any, as plaintext isn't really triggering much, but it may show
>     something).
> 
>     You can also try running extract with the "-i" option ("in-process"),
>     that skips the IPC and runs all plugins in the same address space.
>     Usually good for diagnostics, and makes execution way simpler. But also
>     likely obscures/hides certain problems.
> 
>     My 2 cents
> 
>     Christian
>

0xE29FC3CC.asc
Description: application/pgp-keys