Re: [Synaptic-devel] Strange segfault

[Panu Matilainen]

On Fri, 16 Apr 2004, Andrey Tarassov wrote:

> 
> Panu Matilainen wrote:
> > On Thu, 2004-04-15 at 20:56, Andrey Tarassov wrote:
> > 
> >>Hi!
> >>
> >>I have encountered a very strange problem with synaptic? Namely, we have 
> >>a repository that may be accessed from both http and ftp using the same 
> >>address and the repository is signed with GPG.
> >>
> >>Now, the problem is that if you press the refresh button on startup 
> >>(after some time this is not even required) and edit repositories, 
> >>pressing refresh again causes synaptic to crash.
> >>
> >>There is only one thing I am editing in repositories: ftp -> http and 
> >>vice versa. Nothing else. I also encountered this problem before with 
> >>some other changes, but know I have found a possibility to reproduce the 
> >>bug easily.
> >>
> >>I looked deeper into the problem and tried to trace the problem with 
> >>gdb. I found that the segfault appeared in pkgAcqIndexRel::Done(), but 
> >>could not find why it was happening as everything seemed to be ok.
> >>
> >>Have you heard about such a problem and does anyone know about how to 
> >>solve it?
> >>
> >>FYI, I am using synaptic-0.48.1 and apt-rpm 0.15.5cnc6.
> > 
> > 
> > That sounds a whole lot like a longstanding problem I'm seeing with
> > apt-shell together with signed repositories:
> > 
> > When running "update" from apt-shell, by the time it gets to 
> > pkgAcqIndexRel::Done(), Repository->Vendor->FingerPrint has turned into 
> > garbage, causing the segfault on comparison to Release fingerprint.
> > Actually most of the Vendor stucture gets overwritten by something, it
> > seems consistent that the corruption begins at the last character of
> > Vendor->Description (but not 100% sure about that), anyway the entire
> > Vendor Repository->Vendor struct get garbled useless.
> > 
> > Doesn't happen with synaptic in my environment but it sounds all too
> > familiar - can you post the entire backtrace of the crash so we can see
> > whether it's the same problem or something else?
> 
> Yes, sure. Here we go:
> 
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread -150645536 (LWP 22413)]
> pkgAcqIndexRel::Done (this=0x908b960, Message=Cannot access memory at 
> address 0x0
> ) at basic_string.h:257
> 257           { return &((reinterpret_cast<_Rep*> (_M_data()))[-1]); }
> (gdb) bt
> #0  pkgAcqIndexRel::Done (this=0x908b960, Message=Cannot access memory 
> at address 0x0

Hmm, that's not quite the same as my problem though it happens at the 
same spot (pkgAcqIndexRel::Done): in my case the Message is ok, it's the 
Repository entry which gets garbled.

[snip] 
> In general I believe that you described the situation quite correctly. 

..or maybe not :-/ Does it only happen if you have repository 
authentication on? If that's the case it's probably same issue in a 
slightly different disguise. FWIW I'm not able to reproduce the crash in 
synaptic (same versions used) no matter what I do.

> Do you know if the apt-rpm developers are aware about the problem?

I'm one of them :) Gustavo Niemeyer, who's the main apt-rpm developer, 
also knows about the issue but hasn't commented so I assume that either he 
can't reproduce the problem or is just as mystified as I am.

> 
> I will try to look into this problem deeper, but unfortunately I do not 
> have that much time for this :-(

Likewise - apt-shell not working hasn't been enough of a showstopper to 
put some serious effort into the problem since apt-get and synaptic have 
been working for me but...

        - Panu -