| Re: [AD] broken dat and grabber? |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
Peter Wang <tjaden@xxxxxxxxxx> wrote:
> Peter Wang <tjaden@xxxxxxxxxx> wrote:
> >
> > I got that too with 3.9.34 WIP. I assume it's the bug reported by
> > Emmanuel, Grzegorz and Angelo. Quick fix: use debug library or
> > disable asm.
>
> The problem disappeared when I compiled `blit.c' manually without
> the `-funroll-loops' option, and the rest of the library as usual.
> We could be looking at a gcc bug. Could someone else who got bit by
> this bug try this out?
>
> For me, "gcc -v" shows:
>
> Reading specs from /usr/lib/gcc-lib/i386-slackware-linux/2.95.2/specs
> gcc version 2.95.2 19991024 (release)
>
> I guess I didn't have the problem before because I had egcs 1.1.2.
to bad that happens in a stable release. now we have encountered two gcc
optimalization bugs.
if somebody feels like submitting a bugreport they could send one to Bernd
Schmidt <bernds@xxxxxxxxxx>.
i have been in touch with him mentioning the bug i found in gcc 2.95.3 but
unfortuanetly i had already throw that mandrake 7.2 CD out the window so i
wasn't able to give him much useful info.
useful bugreport recipy:
a preprocessed version of the code that is mis-optimized.
an assembly version of the same thing (optimized build)
an assembly version of the same thing (debug build / non-optimized)
when i say assembly i mean what gcc produces, gdb disassembly will be pretty
useless here but a standard stack/register dump and backtrace would perhaps
be nice...
-henrik