Re: [PATCH] Convert __builtin_expect to __glibc_(un)likely throughout elf/.

[Zack Weinberg <zackw at panix dot com>]

On Mon, Jun 4, 2018 at 11:33 AM, Zack Weinberg <zackw@panix.com> wrote:
> On Mon, Jun 4, 2018 at 10:36 AM, Joseph Myers <joseph@codesourcery.com> wrote:
>> On Sat, 2 Jun 2018, Zack Weinberg wrote:
>>
>>> I would appreciate someone checking over this patch very carefully
>>> indeed.  In a whole bunch of places, __builtin_expect was used in a
>>> way that did not map mechanically to the newer convention.  The test
>>> suite caught one mistake; I'm not confident I didn't make any more.
>>
>> I think this means there is a case for splitting this patch up into a part
>> that doesn't change installed stripped shared libraries (on one platform -
>> so at least the sense of comparisons etc. is unchanged on that platform
>> after the patch, even if likely/unlikely could be wrong sometimes without
>> being detected by such a comparison of installed stripped shared
>> libraries), and a part that does (presumably because the mapping isn't
>> purely mechanical so compiler IR changes enough to result in different
>> code generated), with the latter needing more careful review that no
>> semantics have changed.
>
> I do not have time to do that in this release cycle.  I think my scanf
> patches are much more important and I intend to devote what little
> time I do have for glibc hacking to them.

I have pushed this patch as the branch
zack/elf-builtin-expect-conversion.  I will not be working on it any
more for the foreseeable future but anyone is welcome to pick up the
ball.

zw