[mythtv] mythfilldatabas[16844]: segfault at a0 ip 00007fbc0dd7a8e2 sp 00007fbbfd4ec090 error 4 in libmythdb-0.24.so.0.24.0[7fbc0dcce000+13d000]

[James Courtier-Dutton]

On 30 October 2010 20:58, Michael T. Dean <mtdean at thirdcontact.com> wrote:
>  On 10/30/2010 03:14 PM, James Courtier-Dutton wrote:
>>
>> I am trying to work out why this is happening.
>> The core dump gives this:
>> Program terminated with signal 11, Segmentation fault.
>> #0  0x00007fbc0dd7a8e2 in MythCoreContext::ClearSettingsCache
>> (this=0xdcab60, myKey=...) at mythcorecontext.cpp:629
>> 629         d->m_database->ClearSettingsCache(myKey);
>> (gdb) bt
>> #0  0x00007fbc0dd7a8e2 in MythCoreContext::ClearSettingsCache
>> (this=0xdcab60, myKey=...) at mythcorecontext.cpp:629
>> #1  0x00007fbc0dd7cb45 in MythCoreContext::readyRead (this=0xdcab60,
>> sock=0xdfa140) at mythcorecontext.cpp:890
>> #2  0x00007fbc0dd1f80c in MythSocketThread::ReadyToBeRead
>> (this=0xd92720, sock=0xdfa140) at mythsocketthread.cpp:165
>> #3  0x00007fbc0dd2234d in MythSocketThread::run (this=0xd92720) at
>> mythsocketthread.cpp:352
>> #4  0x00007fbc07e9827e in ?? () from /usr/lib/libQtCore.so.4
>> #5  0x00007fbc097f6953 in ?? () from /usr/lib/nvidia-current/libGL.so.1
>> #6  0x00007fbc09cdb971 in start_thread (arg=<value optimised out>) at
>> pthread_create.c:304
>> #7  0x00007fbc073e791d in clone () at
>> ../sysdeps/unix/sysv/linux/x86_64/clone.S:112
>> #8  0x0000000000000000 in ?? ()
>
> In which app is this happening?  What's occurring when it happens?  Is it
> just:
>
> http://svn.mythtv.org/trac/ticket/7714
>
>> And further analysis means that the segfault is due to d being NULL.
>> But d is created in the constructor.
>> MythCoreContext::MythCoreContext(const QString&binversion,
>>                                  QObject *guiContext)
>>     : d(NULL)
>> {
>>     d = new MythCoreContextPrivate(this, binversion, guiContext);
>> }
>>
>> So, I conclude that this is due to out of memory.
>> The problem I have, is what is the correct fix.
>> When the new call fails, what is the correct error path?
>> One option is just to do this:
>>
>> void MythCoreContext::ClearSettingsCache(const QString&myKey)
>> {
>>     if (d)        //<- add this line.
>>        d->m_database->ClearSettingsCache(myKey);
>> }
>
> Well, if we have no MythContextPrivate, the whole app will fail, so not a
> lot you can do--especially if you're out of memory.  Fail, exit, ...
>
> If m_database is NULL, there may more you can do.
>
> If it's #7714, the proper fix is a bit complex, and we'll probably just
> include http://svn.mythtv.org/trac/changeset/24291 in 0.24 (like we did in
> 0.23) since we don't have time to do (nor want the instability caused by the
> changes required for) the proper fix before release.
>

It is similar to #7714, but I am using the latest SVN as of yesterday,
so the bug fix in #7714 has already been applied.
If the problem I am seeing is a race condition after the object has
been destroyed, then I agree the real fix is probably more
complicated.
If the problem is a simple as out-of-mem then throwing an exception is
probably the only sensible solution.
I will try adding the "if (d)" and see if it hides the problem of the
race condition.