Backporting patch for PR 27291 (Regression fix for PR 28221 on 4.1)
Wed Aug 2 06:45:00 GMT 2006
On Tue, 1 Aug 2006, Steve Ellcey wrote:
> The fix for the 4.1 regression bug PR 28221 is to backport the patch for
> PR 27291 to the 4.1 branch. But in order to do this (or at least to
> make it easy) I want to put the following patch from Nathan Sidwell on
> the 4.1 branch first:
> Then the patch for PR 27291 will install cleanly. Both patches have
> been tested with the 4.1 branch on i486-linux-gnu and on ia64 HP-UX and
> Linux but since the first patch isn't technically a regression fix I
> thought I would see if there are any objections to me putting both
> patches on the 4.1 branch.
> Are there any objections?
My opinion is that Nathan's change is small and safe, perhaps even
obvious, and reasonable to backport as part of a fix to PR28221.
In fact, given that every line changed by Nathan's patch is
subsequently modified by Zdenek's fix for PR tree-optimization/27291,
I think this can/should be considered a straight backport of that
latter fix, and pre-approved under the current guidelines. Having
used the "completely subsumed" argument to motivate a backport, I also
think it's polite to acknowledge Nathan's contribution, even though
the final code is the ultimately the same.
2006-08-02 Steve Ellcey <firstname.lastname@example.org>
Backport from mainline.
2005-11-30 Nathan Sidwell <email@example.com>
* loop-doloop.c (add_test): Only add jump notes if we did emit a
2006-05-01 Zdenek Dvorak <firstname.lastname@example.org>
* loop-doloop.c (add_test, doloop_modify): Handle the case
condition is folded to a constant.
Hopefully others agree with my interpretation of GCC protocol and
etiquette. I'm sure Mark will let me/us know if I'm not inline with
Thanks for fixing this.
More information about the Gcc-patches