This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 0/4] explicit_bzero, this time for sure
On 08/18/2016 02:18 PM, Florian Weimer wrote:
> On 08/18/2016 06:29 PM, Zack Weinberg wrote:
>> On 08/17/2016 06:20 PM, Manuel López-Ibáñez wrote:
>>> On 17/08/16 18:19, Zack Weinberg wrote:
>>>> anything I would like to hear about it. If anyone has any ideas for
>>>> how to write a test that will start failing if a compiler ever learns
>>>> to "see through" __glibc_read_memory, I would like to hear about that,
>>>> too. (I can imagine a way to do it with gcc's scan-assembler tests,
>>>> but we don't have those here.)
>>>
>>> Perhaps? https://plus.google.com/+MatthewDempsky/posts/KQHFBouxurX
>>
>> Ingenious! Done - patch 5/4 attached to this message, and also included
>> in zack/explicit-bzero. I would fold this into patch 2 for landing, but
>> it may be easier to review this way.
>
> GCC scans the generated assembly for such things. It seems more
> reliable, particularly if we retain the explicit_bzero symbol.
Unlike GCC we don't have any infrastructure for doing that, though, and
if we *don't* retain the explicit_bzero symbol (see other message) we
would have to detect all possible inline memset constructs, which
doesn't sound fun.
zw