[vorbis-dev] CVS vorbis, fatal cc1

[Segher Boessenkool]

> Signal 11 in gcc are signs of bad hardware, especially memory, most of the
> time.  You might want to run memtest86 on your PC to be sure.

I wish people would unlearn this "sig11 == hardware problem" as fast and
thoroughly as they learned it.  With any somewhat-recent gcc, a sig11
(and especially repeatable ones!) are almost always a gcc bug.

Also, I wish people would learn to look at the bug database before reporting
a bug.  This one has come up so often already it's not funny anymore.

Also, I wish this would go in the FAQ; alternatively, if it is already in
there, I wish people would read the FAQ.

Also, I wish the libvorbis configure script would autodetect this problem.

Also, I want a pony.

<p>Anyway, the exact problem is:  gcc-2.95.x -mno-ieee-fp -march=i686 segfaults
on compiling about any floating point comparison.  There happen to be a lot
of those in libvorbis, and not many in other packages, and there certainly
are very few packages that default to -mno-ieee-fp on x86.  So we get hit
by this and almost no other package does.

As Michael said, changing -mno-ieee-fp -march=i686 to -mno-ieee-fp
-march=i586 -mcpu=i686 generates the fastest alternative code for
libvorbis.  This is faster code than any gcc-3.x makes for it, btw
(as far as i know).

<p>Have fun,

Segher

<p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.