[mythtv] glibc detected invalid pointer

Anduin Withers awithers at anduin.com
Thu Oct 5 20:53:33 UTC 2006


> Works now, thanks a lot for helping me out! Is this due to some bug in
> ATI's proprietary drivers or in Myth?

Personally? I blame gcc, then ATI (only some), then I start picking lesser
deities at random.

The MythVideo changes were born under a bad sign. I've yet to determine
exactly what was in retrograde to so curse its existence (I have a hunch
about Pluto but the recent downgrade has me now questioning whether I put
too much faith in astrology).

The DRI chunk of the ATI driver uses C++ (this is why I spread the blame all
the way to ATI) and so uses the libstdc++ library (specifically .so.5).
MythTV, compiled on modern distributions (those using gcc >= 3.4), will be
using libstdc++ (specifically .so.6). This should work fine, the fact is
that it does not (though a good deal of work was done getting is just close
enough to make people think/assume it does).

MythVideo is fool enough to use a std::vector<int>, an act so heinous that
only strange crashiness is a fitting punishment. The number of random
applications that one runs that use C++ -- let alone C++, std::vector<int>
and have the ATI DRI module loaded -- is likely to be somewhat small.

Now for some good news: I will be checking in a visibility patch this
weekend that will cause a bit of bellyaching about how recompiles take
for...ever and about what an unnecessary thing it is to be spending time on.
That isn't the good news, the good news is that with any luck (see paragraph
two) it may accidentally fix this.

-- 
Anduin Withers



More information about the mythtv-dev mailing list