Re: glibc 2.23 --- Hard freeze starting

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

On 01/18/2016 03:25 PM, Adhemerval Zanella wrote:
> Maintainers,
> 
> We are now in hard freeze as stated in previous calendar planning [1].
> As before, please keep your commits minimal, documentation changes only,
> simple bug fixes, and no translatable strings.
> 
> Please start test your machines, refresh the ULP files and update the 
> release page [2].
> 
> Current I see that we have 4 blockers right now:
> 
> 1. Add tunables framework: Siddhesh sent a v2 patchset 01/16 without
>    further comments.

I'm OK with this, and we all agree it's just the non-public internals
part of the patch.
 
> 2. Add strlcpy/strlcat: from latest messages I noted Paul has asked to
>    delay it to 2.24 for further discussions.

The status of this work needs a more detailed summary by Florian.

What's blocking adding these functions?

> 3. Fix ABI for external copies of string function inlines on AArch64:
>    there are two patches listed in the wiki page. We need to decide which
>    version to use and narrow down the remaining ABI issues if any.

OK.
 
> 4. [BZ #19329] Fix race between tls allocation at thread creation and dlopen:
>    Szabolcs Nagy is confident it won't add regressions and I saw Carlos
>    only asked for some documentation updates.

This is a rats nest.

My professional opinion is that we should not delay 2.23 for something that
is sufficiently complex to require multiple senior people to review.

I do not doubt that Szabolcs can deliver a solution, but we need a good review
by other senior people which will take time.

Again, I think this has to be delayed until 2.24. We have lived with this bug
for quite a while now. Papering over the problem may introduce other races we
don't know about.

However, I will attempt to review the minimal patch and see what we can get
into 2.23, but I don't think we should delay if we can't get good review
coverage by the deadline.

Cheers,
Carlos.