GDB 7.12 release - 2016-09-21 update?

GDB 7.12 release - 2016-09-21 update?

[Joel Brobecker]

Hello everyone,

Sorry for being radio-silent the last few weeks. It's been a combination
of a crazy traveling schedule, and that thing we call "work"...

Anyways, I reviewed the list of PRs marked as targetting 7.12,
as well as the items listed in our 7.12 release page
(https://sourceware.org/gdb/wiki/GDB_7.12_Release).

It looks like we may have a ways to go before being able to release.
Here are my thoughts on each of them. What do you guys think?

PR 20395: -var-update gives incorrect register values for frames 1 and above

        Assigned to: Don Breazeal
        Status: Patch proposed, but not sure it is good.

        Proposed as blocker for 7.12 because incorrect data is returned.
        However, not sure whether I agree or not yet. I think a better
        question to ask would be: Is it a regression compared to 7.11?

PR 20448: target descriptions: bitfields vs flags backward/forward incompatible change

        Assigned to: *** UNASSIGNED ***
        Status: ???

        This is an actual regression, so I agree we should try to get
        this fixed for 7.12.

        Is anyone working on this?

PR 20627: GDBserver can not be closed

        Assigned to: *** UNASSIGNED ***
        Status: ???

        Not sure if this is a regression or not, and whether it is
        actually critical for 7.12. If there is no data loss, perhaps
        we can just document the workaround.

        Yao?

PR 20487: Thread sync broken between mi and cli with new-ui command

        Assigned to: Antoine Tremblay
        Status: Fix in progress

        Not sure if this is really blocking for 7.12 either as it
        relates to a new feature. But since we seem to be close to
        a resolution, and this makes the new feature more usable,
        perhaps we can try to have it in 7.12. Otherwise, perhaps
        release 7.12 with this limitation in the release notes,
        and fix it for 7.12.1???

PR 11094: cannot reexecute programs using the GDB JIT interface in shared libs

        Assigned to: Pedro Alves
        Status: Fix posted.

        I don't think it's the reported issue itself that was the cause
        of making this PR blocking for 7.12, but rather the SEGV that
        was detected after the patch was pushed in. The fix was posted,
        so I suspect we can have a quick resolution for this one.

PR 20609 gdb --pid $(pidof qemu-system-x86_64) stopped working with gdb 7.11.1

        Assigned to: *** UNASSIGNED ***
        Status: ???

        As hinted by the subject, this is a real regression, and
        the problem description says that this leads to a non-functional
        session.

        Is anyone working on this?

Thanks!
-- 
Joel

Re: GDB 7.12 release - 2016-09-21 update?

[Simon Marchi]

On 2016-09-21 13:16, Joel Brobecker wrote:
> Hello everyone,
> 
> Sorry for being radio-silent the last few weeks. It's been a 
> combination
> of a crazy traveling schedule, and that thing we call "work"...
> 
> Anyways, I reviewed the list of PRs marked as targetting 7.12,
> as well as the items listed in our 7.12 release page
> (https://sourceware.org/gdb/wiki/GDB_7.12_Release).
> 
> It looks like we may have a ways to go before being able to release.
> Here are my thoughts on each of them. What do you guys think?

Thanks for doing this, tell us if you need any help for anything.

> PR 20487: Thread sync broken between mi and cli with new-ui command
> 
>         Assigned to: Antoine Tremblay
>         Status: Fix in progress
> 
>         Not sure if this is really blocking for 7.12 either as it
>         relates to a new feature. But since we seem to be close to
>         a resolution, and this makes the new feature more usable,
>         perhaps we can try to have it in 7.12. Otherwise, perhaps
>         release 7.12 with this limitation in the release notes,
>         and fix it for 7.12.1???

I took over this one.  As you said, since it's not a regression, it's 
not critical.  But as there are still other issues to be solved before 
the release, I'll continue to aim to have it in 7.12.  If it happened to 
be the last remaining open issue, it should certainly not block the 
release.

Simon

Re: GDB 7.12 release - 2016-09-21 update?

[Simon Marchi]

On 16-09-21 01:46 PM, Simon Marchi wrote:
>> PR 20487: Thread sync broken between mi and cli with new-ui command
>>
>>         Assigned to: Antoine Tremblay
>>         Status: Fix in progress
>>
>>         Not sure if this is really blocking for 7.12 either as it
>>         relates to a new feature. But since we seem to be close to
>>         a resolution, and this makes the new feature more usable,
>>         perhaps we can try to have it in 7.12. Otherwise, perhaps
>>         release 7.12 with this limitation in the release notes,
>>         and fix it for 7.12.1???
> 
> I took over this one.  As you said, since it's not a regression, it's 
> not critical.  But as there are still other issues to be solved before 
> the release, I'll continue to aim to have it in 7.12.  If it happened to 
> be the last remaining open issue, it should certainly not block the 
> release.

With the two patches I just pushed to the branch (as well as to master),
you can consider this issue fixed, unless I broke something I am not yet
aware of, of course.

Simon

Re: GDB 7.12 release - 2016-09-21 update?

[Don Breazeal]

On 9/21/2016 10:16 AM, Joel Brobecker wrote:

> 
> PR 20395: -var-update gives incorrect register values for frames 1 and above
> 
>         Assigned to: Don Breazeal
>         Status: Patch proposed, but not sure it is good.
> 
>         Proposed as blocker for 7.12 because incorrect data is returned.
>         However, not sure whether I agree or not yet. I think a better
>         question to ask would be: Is it a regression compared to 7.11?
> 
Hi Joel,
This problem was present in 7.11.
Thanks
--Don

Re: GDB 7.12 release - 2016-09-21 update?

[Yao Qi]

On Wed, Sep 21, 2016 at 6:16 PM, Joel Brobecker <brobecker@adacore.com> wrote:

>
> PR 20627: GDBserver can not be closed
>
>         Assigned to: *** UNASSIGNED ***
>         Status: ???
>
>         Not sure if this is a regression or not, and whether it is
>         actually critical for 7.12. If there is no data loss, perhaps
>         we can just document the workaround.
>
>         Yao?
>

There is no such problem on 7.11.1.

Since the GDBserver can not be closed, GDB can not quit until it gets
packet error.  (I start GDB in one window and start GDBserver in the
other.)

  (gdb) quit
  A debugging session is active.

  Inferior 1 [process 14736] will be killed.

  Quit anyway? (y or n) y
  Ignoring packet error, continuing...

In my real remote testing, GDB can quit and GDBserver process leaves
on the target machine.  There are a lot of GDBserver processes left after
running GDB testsuite.

I am looking into this bug...

-- 
Yao (齐尧)

Re: GDB 7.12 release - 2016-09-21 update?

[Jan Kratochvil]

On Wed, 21 Sep 2016 19:16:40 +0200, Joel Brobecker wrote:
> PR 20609 gdb --pid $(pidof qemu-system-x86_64) stopped working with gdb 7.11.1
> 
>         Assigned to: *** UNASSIGNED ***
>         Status: ???
> 
>         As hinted by the subject, this is a real regression, and
>         the problem description says that this leads to a non-functional
>         session.
> 
>         Is anyone working on this?

Looking at it.


Jan

Re: GDB 7.12 release - 2016-09-21 update?

[Pedro Alves]

On 09/21/2016 06:16 PM, Joel Brobecker wrote:
> PR 20448: target descriptions: bitfields vs flags backward/forward incompatible change
> 
>         Assigned to: *** UNASSIGNED ***
>         Status: ???
> 
>         This is an actual regression, so I agree we should try to get
>         this fixed for 7.12.
> 
>         Is anyone working on this?

Doug had fixed this already, but the fix somehow never made it to
master.  Fix pushed to master now, though I have some questions open.
Most likely it's just me being confused, though I'd like to
double-check that tdescs are behaving as intended, since we'll be
stuck with whatever we decide:

 https://sourceware.org/ml/gdb-patches/2016-10/msg00113.html
 https://sourceware.org/ml/gdb-patches/2016-10/msg00114.html

This is the only PR left with target-milestone==7.12.

Thanks,
Pedro Alves

Re: GDB 7.12 release - 2016-09-21 update?

[Pedro Alves]

On 10/06/2016 01:12 PM, Pedro Alves wrote:

>  https://sourceware.org/ml/gdb-patches/2016-10/msg00113.html
>  https://sourceware.org/ml/gdb-patches/2016-10/msg00114.html
> 
> This is the only PR left with target-milestone==7.12.

... and ... now closed.

The buildbots are still catching up on the latest commits, but
so far so good.  Actually, there's a --disable-build-with-cxx
build problem on the 7.12 branch right now.  We should fix that
before we release.

Thanks,
Pedro Alves

Re: GDB 7.12 release - 2016-09-21 update?

[Maciej W. Rozycki]

On Thu, 6 Oct 2016, Pedro Alves wrote:

> >  https://sourceware.org/ml/gdb-patches/2016-10/msg00113.html
> >  https://sourceware.org/ml/gdb-patches/2016-10/msg00114.html
> > 
> > This is the only PR left with target-milestone==7.12.
> 
> ... and ... now closed.
> 
> The buildbots are still catching up on the latest commits, but
> so far so good.  Actually, there's a --disable-build-with-cxx
> build problem on the 7.12 branch right now.  We should fix that
> before we release.

 FYI, I have backported the three changes I made on trunk today, including 
the MIPS FCR fix I mentioned the other day in particular, and also its 
preparatory cosmetic update and an obvious fix for gdb.base/solib-disc.exp 
to work at all.

  Maciej

Re: GDB 7.12 release - 2016-09-21 update?

[Joel Brobecker]

> >  https://sourceware.org/ml/gdb-patches/2016-10/msg00113.html
> >  https://sourceware.org/ml/gdb-patches/2016-10/msg00114.html
> > 
> > This is the only PR left with target-milestone==7.12.
> 
> ... and ... now closed.

Good! So, let's hope that was the last one for 7.12.

> The buildbots are still catching up on the latest commits, but
> so far so good.  Actually, there's a --disable-build-with-cxx
> build problem on the 7.12 branch right now.  We should fix that
> before we release.

Let me know when done, and I will produce the release.

Thanks, all!
-- 
Joel

Re: GDB 7.12 release - 2016-09-21 update?

[Pedro Alves]

On 10/06/2016 05:41 PM, Joel Brobecker wrote:

>> The buildbots are still catching up on the latest commits, but
>> so far so good.  Actually, there's a --disable-build-with-cxx
>> build problem on the 7.12 branch right now.  We should fix that
>> before we release.
> 
> Let me know when done, and I will produce the release.
> 
> Thanks, all!

I looked around at the buildbot test results and nothing
looked suspicious.  I think we're good to go.

Thanks,
Pedro Alves

Re: GDB 7.12 release - 2016-09-21 update?

[Joel Brobecker]

> I looked around at the buildbot test results and nothing
> looked suspicious.  I think we're good to go.

Thanks, Pedro. Give me a few more minutes to get my gloves on,
and let's DO this! ;-)

A _big_ thank you to everyone who's been taking care of of all
the known issues, and making sure they got fixed. This really
helped me be less hands-on than usual.

-- 
Joel