Re: [PATCH v3] explicit_bzero yet again

["Carlos O'Donell" <carlos at redhat dot com>]

On 01/11/2016 10:15 AM, Zack Weinberg wrote:
> On 12/21/2015 03:32 PM, Zack Weinberg wrote:
>> On Mon, Dec 7, 2015 at 9:05 AM, Zack Weinberg <zackw@panix.com> wrote:
>>> On 11/19/2015 09:46 AM, Zack Weinberg wrote:
>>>> I've updated the explicit_bero patches for the change to how .abilist
>>>> files work and the reorganization of x86.  I've also split the
>>>> controversial string2.h optimization into its own patch so it can be
>>>> discussed separately from the merits of the interface.
>>>
>>> Ping.  The patches at
>>> <https://sourceware.org/ml/libc-alpha/2015-11/msg00467.html> are
>>> awaiting review.
>>
>> Ping.
> 
> Ping.
> 
> My hacking time is very limited right now.  With the freeze looming, I
> would appreciate a definitive yes-or-no answer to whether this will be
> accepted for the next release (possibly with some concrete list of
> changes) so I know whether I need to find time in the near future.

I'd nack this for 2.23, and review again in 2.24.

The clear points from this discussion are:

(a) glibc does support static linking.

    If you see anywhere that we are saying we don't support static
    linking, please point it out and I will correct this.

    We will support static linking fully, but with caveats that
    allow the user to accept the problems that come with having
    two namespaces in their application (the static one and the
    dynamic one).

(b) LTO problems.

    You leave an open question to Rich about what kind of asm
    barrier would be best to use today (given that no better
    barrier exists). The fact that neither you nor Rich have
    decided yet what should be used means we need more discussion.

    If there isn't already a gcc bug filed for creating this barrier
    we should file one now.

All of this means to me that this should be moved to 2.24.

Cheers,
Carlos.