[mythtv] Deciphering backtrace of loop/segfault

[Robert McNamara]

On Tue, Apr 6, 2010 at 10:08 AM, Kevin Kuphal <kkuphal at gmail.com> wrote:
> On Tue, Apr 6, 2010 at 12:02 PM, Robert McNamara
> <robert.mcnamara at gmail.com> wrote:
>> On Tue, Apr 6, 2010 at 9:57 AM, Michael T. Dean <mtdean at thirdcontact.com> wrote:
>>> On 04/06/2010 08:27 AM, Kevin Kuphal wrote:
>>>>
>>>> I have a friend who updated to 0.23-release tag from SVN and is
>>>> experiencing the situation where his frontend segfaults when watching
>>>> a recording.  I already had him clear out everything myth, distclean,
>>>> recompile, etc. etc.  He compiled a debug copy of Myth and captured
>>>> this backtrace.  Under gdb using the instructions on mythtv.org, the
>>>> frontend doesn't segfault but sits at the "Please wait" screen.   Here
>>>> is the results:
>>>>
>>>> http://pastebin.com/KXuaJGBY
>>>>
>>>> Any help would be greatly appreciated.
>>>>
>>>
>>> Kevin,
>>>
>>> Though the backtrace doesn't tell me much (perhaps others can see more),
>>> this sounds a lot like the the known race condition that "inspired"
>>> http://svn.mythtv.org/trac/changeset/22350 as a workaround.  The real fix is
>>> much more complex, and (TTBOMK), Mark K. will likely tackle it in the
>>> libmythui-osd branch in order to allow the UI code and video code to be used
>>> on the display at the same time.
>>>
>>> Does changing the 50ms sleep to some other value change the results?  If
>>> not, I have another (more invasive) hack of a patch that changes timing more
>>> significantly and can give you a copy to test as a system-specific
>>> workaround.
>>>
>>> Mike
>>
>> This backtrace looks like someone has libmyth* junk left over from an
>> old install.  did you have him rm /usr/lib/libmyth* and/or
>> /usr/local/lib/libmyth*?
>
> I can have him clean that stuff out (I thought we had) but I know that
> when I do a "ldd mythfrontend" that it only shows it linking against
> the 0.23 libraries in the /usr/lib64 folder per our ./configure
>
> Kevin

For what it's worth I tend to think that it's actually both the above
and what Mike mentioned, I just know we've seen this same backtrace a
couple times and in each case it was crufty leftovers that needed
cleaning out in the lib dirs.  But one thing at a time, I guess.

Robert