Re: Code freeze for glibc-2.19

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

On 01/01/2014 07:51 AM, Allan McRae wrote:
> There is still a few patches listed in the "Release blockers" section of
> the wiki page [1].  It looks like the "Make TLS variables thread-safe"
> patch needed changes and is large so can target 2.20.

It is my opinion that we can get this into 2.19 right now. This
will accelerate the adoption and continued testing of signal-safe
TLS variable accesses. Further it will be a stepping stone towards
a GNU2 TLS signal-safe implementation.

Paul, and Andrew please correct me if I'm wrong but the TLS thread-safe
variables work is almost complete and in place for 2.19 (3 of 4 split
patches checked in).

The work to make TLS variables thread-safe was split into 4 patches.

The following 3 patches were reviewed and checked in:

(1) Add signal mask control utility functions.
    - This was reviewed and checked in on 2013-12-18.

(2) Add internal signal safe malloc implementation.
    - This was reviewed and checked in on 2013-12-18.

(3) Factor out _dl_clear_dtv
    - This was reviewed and checked in on 2013-12-18.

There is one last patch to review, which has already received review
by Torvald Riegel, Alex Oliva and myself. Andrew Hunter has provided
a new patch incorporating all suggestions.

(4) Async signal-safe TLS access
    - Reviewed by mew today 2014-01-01 (version 5 of patch)
    - Has a test case.
    - Contains P&C comments about barriers.

Given my review I think it's ready to checkin for 2.19.

However, the final call is Allan's.

Allan, I've tried to give you enough background here for you
to make a call about this patch. That's what I would have
expected from anyone else trying to get a feature into 2.19
at or near the deadline :-)

Comments?

Cheers,
Carlos.