Re: [Gnash] [backtrace included] Gnash segfaults at NPR site

[Rob Savoye]

strk wrote:

>>> This might be because I have an i810 GPU in my computer, and I didn't
>>> pass the --enable-i810-lod-bias option to configure. I recompiled Gnash
>>> doing that, and I got a failed assertion instead of the segfault:

  I've been out camping for a week, and catching up to 1800+ email
message, so sorry this is late.

  The default DRI implementation for the i810 and i815 (at least on
Gentoo) is seriously broken. It has a bug that prevents memory blocks
that have been free()'d, never get returned to the kernel, so if you run
an OpenGL program 24/7 for days, it eats up all the available memory
till the machine crashes...

  Supposedly the LOD bias problem has been fixed for about 10 months in
libMESA, so this option should be unnecessary.

        - rob -