[mythtv] Backtrace: Goom segfaults
Isaac Richards
ijr at case.edu
Mon Feb 21 07:25:54 UTC 2005
On Sunday 20 February 2005 08:04 pm, Ed Wildgoose wrote:
> Carl Alexander wrote:
> > Yes, that may fix the crash at that point, but the question is *why*
> > is it going wrong? What is the design or implementation inconsistency
> > that is causing this? I haven't groked enough of the code to answer that.
> >
> > has anyone valgrind'ed the code? Maybe a UMR or a over read is filling
> > some of the variables with garbage data.
>
> No, I think the data looks sane. The values going into that expression
> are like 160 or something (look at your backtrace). The answer just
> goes slightly negative instead of capping at zero (I think).
>
> Probably there is some kind of rounding error or something subtle which
> is just tripping it over the threshold (at least I think that's what's
> going on...)
Should be easy to figure out what those values are and trap them at a higher
level, so all 3 functions can be fixed in the same place.. Just have it
print out the values it's using to calculate px/py if either go below 0.
Isaac
More information about the mythtv-dev
mailing list