Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Tobias Burnus]

Regarding GCC 4.9:

Joel Brobecker wrote:
> It would be good to get a sense of when people think gcc-4.9 is going
> to come out. Last I looked, the branch hadn't been created yet.
> Perhaps, one option is to schedule 7.8 to coincide with gcc-4.9
> (or slightly before, if we'd like to).

Well, since a couple of years, the GCC branching happens when the
first release candidate is build. The actual release is then one or
two weeks later - depending how many P1 bugs or other issues are
found. Regarding the release time: In the last few years, GCC was
usually released in March, April or May.


In case of GCC 4.8.0: RC1 plus branching March 16, 4.8.0 release
March 21 (announced: March 22). See also
http://gcc.gnu.org/develop.html#timeline

Regarding 4.9: My guess would be that it will take a bit longer
this year due to some larger changes late in the game. Last status
report (1 month ago): http://gcc.gnu.org/ml/gcc/2013-12/msg00203.html
See also "Serious regressions" for 4.9 on http://gcc.gnu.org/

Tobias

heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Joel Brobecker]

Hello,

I really was planning on sending this email at the end of last
week, and then forgot. It's been a fairly quiet 3 weeks since
the first pre-release, and there are no longer any known failures
mentioned in the release wiki page, so I am planning on making
the official release tomorrow. Please shout if there is something
I missed :).

Thank you,
-- 
Joel

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Sergio Durigan Junior]

On Tuesday, January 28 2014, Joel Brobecker wrote:

> I really was planning on sending this email at the end of last
> week, and then forgot. It's been a fairly quiet 3 weeks since
> the first pre-release, and there are no longer any known failures
> mentioned in the release wiki page, so I am planning on making
> the official release tomorrow. Please shout if there is something
> I missed :).

Heya!

This is definitely not a blocker, but I'd like to know what are the
status of the patches submitted by Edjunior for PPC64.  There is at
least one of them which solves a false-positive, so for the sake of
avoiding wasting time if someone finds these "failures" in the future,
it might be good to check those in.

Thanks,

-- 
Sergio

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Joel Brobecker]

> This is definitely not a blocker, but I'd like to know what are the
> status of the patches submitted by Edjunior for PPC64.  There is at
> least one of them which solves a false-positive, so for the sake of
> avoiding wasting time if someone finds these "failures" in the future,
> it might be good to check those in.

OK, thanks for the heads up. It doesn't sounds like it would be
blocking indeed, but I'll wait for EdJ's feedback if he can send it
today or tomorrow...

Thanks,
-- 
Joel

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Ulrich Weigand]

Joel Brobecker wrote:

> I really was planning on sending this email at the end of last
> week, and then forgot. It's been a fairly quiet 3 weeks since
> the first pre-release, and there are no longer any known failures
> mentioned in the release wiki page, so I am planning on making
> the official release tomorrow. Please shout if there is something
> I missed :).

You may have noticed the series of patches I just posted to support
the new powerpc64le-linux platform.

I'm certainly aware that this comes in at a very late time, but
given the timing of when powerpc64le-linux support shows up in
the other parts of the core toolchain (binutils 2.24, gcc 4.9,
glibc 2.19) it would be really nice if GDB 7.7 also had support,
to complete the picture ...

On the other hand, it would certainly be preferable to allow for
at least a couple of days before checking the series in to give
people a chance to comment, so it seems unlikely this could be
in place in the branch by tomorrow.

So I guess I'm wondering if you see workable options to still
get powerpc64le-linux support into a release in the near future ...
Maybe we could do a 7.7.1 ? 

Sorry again for the last-minute request; the whole powerpc64le
project has been running on a short timeline, and GDB support
got pushed back by other requirements :-(

Thanks for your support!

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU/Linux compilers and toolchain
  Ulrich.Weigand@de.ibm.com

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Joel Brobecker]

> I'm certainly aware that this comes in at a very late time, but
> given the timing of when powerpc64le-linux support shows up in
> the other parts of the core toolchain (binutils 2.24, gcc 4.9,
> glibc 2.19) it would be really nice if GDB 7.7 also had support,
> to complete the picture ...

Outch.

> On the other hand, it would certainly be preferable to allow for
> at least a couple of days before checking the series in to give
> people a chance to comment, so it seems unlikely this could be
> in place in the branch by tomorrow.

I don't know of any reason that would force us to release tomorrow.
The last 2-3 weeks of pre-release testing having been fairly quiet,
we had no real reason to wait either.

> So I guess I'm wondering if you see workable options to still
> get powerpc64le-linux support into a release in the near future ...
> Maybe we could do a 7.7.1 ?

I'm a bit uncomfortable including support for a new target like that
without more testing, particularly since the changes affects files
already in use by other ports. I had a quick look, and perhaps the
changes can be reviewed well enough to convince ourselves that they
are indeed safe enough for the branch.

Here is what I propose:

  1. We delay the release to Monday (Feb 3rd).

  2. During that time frame, let's try to get the changes reviewed
     by at least one more GM - it might be me.

It would be good to get a sense of when people think gcc-4.9 is going
to come out. Last I looked, the branch hadn't been created yet.
Perhaps, one option is to schedule 7.8 to coincide with gcc-4.9
(or slightly before, if we'd like to).

Cheers,
-- 
Joel

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Ulrich Weigand]

Joel Brobecker wrote:

> > So I guess I'm wondering if you see workable options to still
> > get powerpc64le-linux support into a release in the near future ...
> > Maybe we could do a 7.7.1 ?
> 
> I'm a bit uncomfortable including support for a new target like that
> without more testing, particularly since the changes affects files
> already in use by other ports. I had a quick look, and perhaps the
> changes can be reviewed well enough to convince ourselves that they
> are indeed safe enough for the branch.
> 
> Here is what I propose:
> 
>   1. We delay the release to Monday (Feb 3rd).
> 
>   2. During that time frame, let's try to get the changes reviewed
>      by at least one more GM - it might be me.

I would certainly appreciate that -- thanks!

As to changes in other ports, I think it should be pretty clear
that non-PowerPC platforms cannot really be affected.  (There are
some changes to common code in the 11/11 patch, but that shouldn't
affect any platform that do not define gdbarch_skip_entrypoint.)

As to other PowerPC platforms, I've tested powerpc64-linux quite a
bit -- I would be happy to also run a regression test on a 32-bit
Linux and on AIX.

> It would be good to get a sense of when people think gcc-4.9 is going
> to come out. Last I looked, the branch hadn't been created yet.
> Perhaps, one option is to schedule 7.8 to coincide with gcc-4.9
> (or slightly before, if we'd like to).

I guess that might be OK as a fall-back solution.

Thanks,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU/Linux compilers and toolchain
  Ulrich.Weigand@de.ibm.com

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Joel Brobecker]

> I would certainly appreciate that -- thanks!
> 
> As to changes in other ports, I think it should be pretty clear
> that non-PowerPC platforms cannot really be affected.  (There are
> some changes to common code in the 11/11 patch, but that shouldn't
> affect any platform that do not define gdbarch_skip_entrypoint.)
> 
> As to other PowerPC platforms, I've tested powerpc64-linux quite a
> bit -- I would be happy to also run a regression test on a 32-bit
> Linux and on AIX.

OK, I've had a chance to review the patches, and I am definitely
hesitating. I agree that they can only affect PowerPC. If you feel
very confident about the patches and want to take responsibility,
I will go with your call. Advice from the other GMs would also be
useful.

My preference goes towards getting this code in the master branch
first, release 7.7 without it, and then re-evaluate. If proven after
an observation period to be stable in master, we can elect to backport
the code to 7.7 in time for 7.7.1. But if people prefer, we can also
accelerate the schedule for 7.8.

-- 
Joel

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Mark Kettenis]

> Date: Wed, 29 Jan 2014 12:14:08 +0400
> From: Joel Brobecker <brobecker@adacore.com>
> 
> > I would certainly appreciate that -- thanks!
> > 
> > As to changes in other ports, I think it should be pretty clear
> > that non-PowerPC platforms cannot really be affected.  (There are
> > some changes to common code in the 11/11 patch, but that shouldn't
> > affect any platform that do not define gdbarch_skip_entrypoint.)
> > 
> > As to other PowerPC platforms, I've tested powerpc64-linux quite a
> > bit -- I would be happy to also run a regression test on a 32-bit
> > Linux and on AIX.
> 
> OK, I've had a chance to review the patches, and I am definitely
> hesitating. I agree that they can only affect PowerPC. If you feel
> very confident about the patches and want to take responsibility,
> I will go with your call. Advice from the other GMs would also be
> useful.
> 
> My preference goes towards getting this code in the master branch
> first, release 7.7 without it, and then re-evaluate. If proven after
> an observation period to be stable in master, we can elect to backport
> the code to 7.7 in time for 7.7.1. But if people prefer, we can also
> accelerate the schedule for 7.8.

I'd say that rushing changes like this in just before release is just
bad engineering practice in general.  And in this case I have a hard
time convincing myself that this will not break existing big-endian
PowerPC targets.  This is the sort of stuff that you want to get in
early in a release cycle.  I also think we shouldn't do these
feature-driven release adjustments.  There's just to many external
variables to take them into account.

So I'd say this should go onto master after 7.7 has been released, and
7.8 should be released according to the normal schedule.  If meanwhile
IBM or any of the Linux distros wants to ship a 64-bit wrong-endian
PowerPC toolchain they can just take 7.7 and apply the patches.

Re: heads up: GDB 7.7 official release Wed Jan 29th (tomorrow)

[Ulrich Weigand]

Mark Kettenis wrote:
> > From: Joel Brobecker <brobecker@adacore.com>
> > 
> > OK, I've had a chance to review the patches, and I am definitely
> > hesitating. I agree that they can only affect PowerPC. If you feel
> > very confident about the patches and want to take responsibility,
> > I will go with your call. Advice from the other GMs would also be
> > useful.
> > 
> > My preference goes towards getting this code in the master branch
> > first, release 7.7 without it, and then re-evaluate. If proven after
> > an observation period to be stable in master, we can elect to backport
> > the code to 7.7 in time for 7.7.1. But if people prefer, we can also
> > accelerate the schedule for 7.8.
> 
> I'd say that rushing changes like this in just before release is just
> bad engineering practice in general.  And in this case I have a hard
> time convincing myself that this will not break existing big-endian
> PowerPC targets.  This is the sort of stuff that you want to get in
> early in a release cycle.  I also think we shouldn't do these
> feature-driven release adjustments.  There's just to many external
> variables to take them into account.
> 
> So I'd say this should go onto master after 7.7 has been released, and
> 7.8 should be released according to the normal schedule.  If meanwhile
> IBM or any of the Linux distros wants to ship a 64-bit wrong-endian
> PowerPC toolchain they can just take 7.7 and apply the patches.

Oh, I'm not so much concerned about the distros -- we'll get those sorted
out (and they may need backports to even earlier releases anyway).  It
just seems it would have been nice for users of the FSF releases to have
an official GDB release that works on this new platform in a similar time
frame as the other components of the toolchain become available.

In any case, if the general sentiment is that we shouldn't do this as a
last-minute change at this stage, I can certainly understand and accept that.

Thanks,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU/Linux compilers and toolchain
  Ulrich.Weigand@de.ibm.com