Re: [RFC] any critical patches for a release this weekend?

[Peter Rosin]

On 2014-11-02 11:42, Gary V. Vaughan wrote:
> [[Added back Cc: libtool-list]]
> 
>> On Nov 1, 2014, at 11:11 AM, Richard PALO <address@hidden> wrote:
>>
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Le 27/10/14 23:02, Gary V. Vaughan a écrit :
>> | Hi Richard,
>> |
>> |> I'd like to see two patches committed:
>> |>
>> |> 1. the attached SunOS g++ shared library patch, in pkgsrc, has
>> |> been tested and running in production here since over half a
>> |> year, and I understand Joyent is using it in their local branch.
>> |
>> | Unfortunately, without some more background explanation regarding
>> | the reason why the patch is necessary, what it fixes, and the
>> | rationale thereof (and a ChangeLog entry or similar) I couldn't
>> | evaluate the purpose or potential side-effects of the first patch.
>> | If you could provide those, I'll certainly make the time to
>> | evaluate, and maybe apply if all is well...
>> |
>> | Cheers,
>>
>> Missed your message somehow. Sorry.
> 
> No problem, we still have about a week to untangle things before 2.4.4.
> 
>> This patch snippet is a portion of Peter Rosin's 'g++ and -nostdlib'
>> patch for SunOS.
>>
>> It removes '-nostdlib" from the g++ shared build line and, to avoid
>> duplicate symbols, it doesn't explicitly include '$predep_objects' and
>> '$postdep_objects'.
>>
>> With this patch, g++ shared libraries build like gcc shared libraries,
>> albeit without '-Wl,z text'...
>>
>> This allows finally, for this particular example, that the various
>> '-fstack-protector*' specs build with g++ on SunOS needing:
>> | *link_ssp:
>> |
>> %{fstack-protector|fstack-protector-strong|fstack-protector-all:-lssp_nonshared
>> | -lssp}
> 
> Sorry I didn't word my original reply more carefully.  The stack-protector
> patch was a no-brainer, and as such is included in 2.4.3 already.
> 
> Could you explain the other one of the two attachments you sent originally,
> as I don't have a rationale nor understanding of that one sufficient to make
> me apply it yet :)

That patch appears to be a small part of [1]. There is more info in that
thread, with references to bug reports etc, if you haven't found that already.

Note that the patch was mainly written to spur discussion. I am not ready to
push the whole thing without writing further tests, as indicated in that
thread. I would also be reluctant to pick out one platform and do the change
there, since any regression will likely be present on all platforms.

In short, noone knows the full effects [1]. Sure, it helps for certain cases,
but are there any regressions? Tests needed.

Sadly, I have very limited time to spend on Libtool at the moment...

Cheers,
Peter

[1] http://lists.gnu.org/archive/html/libtool-patches/2013-11/msg00006.html