From: René Scharfe <l.s.r@web.de>
Date: Wed, 03 May 2017 19:48:31 GMT
Subject: Re: [PATCH] cache-tree: reject entries with null sha1
Message-ID: <cddd09e5-8e8a-3dc9-001f-f307d3647619@web.de>
In-Reply-To: <CACsJy8CdjLySD+UEKE2NkbYqEoVOQtvydFBui6UbbnFLo8U1Ug@mail.gmail.com>

Am 03.05.2017 um 11:46 schrieb Duy Nguyen:
> On Tue, May 2, 2017 at 2:22 AM, Jeff King <peff@peff.net> wrote:
>> On Mon, May 01, 2017 at 01:23:28PM +0200, René Scharfe wrote:
>>>
>>> I can only get gcc and clang to call memcpy instead of inlining it by
>>> specifying -fno-builtin.  Do you use that option?  If yes, why?  (Just
>>> curious.)
>>
>> I do my normal edit-compile cycles with -O0 because it's fast, and
>> because it makes debugging much easier.
> 
> Same here. My CFLAGS (without lots of -Wstuff)
> 
> CFLAGS =  -g -O0 -fstack-protector
> 
> Maybe it's -fstack-protector then? This is gcc 4.9.3. I think Gentoo
> does not add any distro-specific patches on this particular version.

gcc 4.9.2 on Debian i386 still inlines memcpy for me with these options.

https://packages.debian.org/jessie/gcc-4.9 links to a 5MB diff, and it
adds these lines to NEWS.gcc (plus changing a whole lot of other files,
of course):

+ - Better inlining of memcpy and memset that is aware of value ranges
+   and produces shorter alignment prologues.

That might be it.

René