[PATCH v2, middle end]: Fix PR 37908, thinko with atomic NAND operation

[Andrew Haley]

Jakub Jelinek wrote:
> On Wed, Oct 29, 2008 at 09:05:09AM +0100, Uros Bizjak wrote:
>> On Tue, Oct 28, 2008 at 11:04 PM, Jakub Jelinek <jakub@redhat.com> wrote:
>> Attached (now for real :) patch corrects __sync nand intrinsics to
>> generate ~(target & val). The patch was tested on i686-pc-linux-gnu. I
>> propose to commit the patch, since all unfixed targets will fail fixed
>> sync tests and after all targets are fixed, we can backport complete
>> patch to release branches and  add the note to release notes.
> 
> IMNSHO we want this in 4.4 only, otherwise people who already use these
> intrinsics have no easy way (except for configure) to find out what semantics
> the builtin has.
> If it is only in 4.4, they can check __GNUC_MAJOR__ and __GNUC_MINOR__.

Yes: this change is risky as it may mysteriously break OS kernels in
difficult to predict ways.  Before we change well-documented gcc behaviour
we should find out who depends on the current behaviour.

Andrew.