Re: [avr-gcc-list] avr-gcc 4.5.3 in Debian/Ubuntu has call-prologues bug

[Dale]

Hi,

There's an update to the Debian package which will be: gcc-avr_4.5.3-4

This version has backported fixes for:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46779
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50289

From Hakan, the Debian maintainer:
"I'm releasing now, it should show up in debian unstable during the day
and propagate to testing after 10 days if no issues show up. How/When
it'll propagate to ubunu I dont know.

This release breaks gcc.c-torture/execute/pr23135.c and
gcc.c-torture/execute/simd-1.c from the gcc testsuit. They are both
dealing with SIMD-support but supposed to pass even if the target does
not have SIMD (they do pass on version 4.5.3-3). But I guess this is a
minor issue..."


> Hi,
> 
> A warning that this version has the bug as described and fixed by Georg
> below. The fix was targeted for 4.6.2
> 
> [avr]: call-prologues saving/restoring global register variables
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50289
> 
> avr-gcc 4.5.3 is in:
> Debian: wheezy(testing) and sid(unstable)
> Ubuntu: oneiric(current release) and precise(next release)
> 
> Compiling without -mcall-prologues appears to be fine.
> 
> I had no problems with 4.3.5 and since this bug can result in erratic
> behaviour (R7 global register is saved/restored by prolog/epilog when it
> shouldn't be), I was wondering whether there are any thoughts on the
> best action.
> 
> Either backport the fix to 4.5.x or move ahead to 4.6.x. Opinions?
> 
> -- 
> 
> Regards
> 
> Dale
> 
> _______________________________________________
> AVR-GCC-list mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/avr-gcc-list
>