GDB 10.2 release (respin) -- 2021-02-28 Update

GDB 10.2 release (respin) -- 2021-02-28 Update

[Joel Brobecker]

Hello,

Here is a quick update on where we are for the 10.2 corrective
release. Only one more to go...

Fixed Since the Previous Update:
--------------------------------

  * [SimonM] <PR gdb/26861>
    Bug 26861 - internal-error: void target_mourn_inferior(ptid_t): Assertion `ptid == inferior_ptid' failed. OS: Mac OSX Catalina

  * [SimonM] <PR gdb/26828>
    SIGSEGV in follow_die_offset dwarf2/read.c:22950

Thank you Simon for your efforts getting those fixed!

Added Since the Last Update:
----------------------------

  < none :) >

Other Ongoing Items:
--------------------

  * [<Simon + Andrew>] <PR backtrace/27147>
    [GNU/Linux, sparc64] GDB is unable to print full stack trace (got "previous frame inner to this frame" errors)
    https://sourceware.org/bugzilla/show_bug.cgi?id=27147

    As the subject says, it's an unwinder issue.  Simon was able
    to reproduce, and tied this issue to a specific commit adding
    multi-target support.

    Simon sent a patch to the PR which is a minimal-impact attempt
    at fixing the issue, which referencing a a larger-scope attempt
    at reducing our dependencies on "current_something" globals...

Not Critical, but Requested:
----------------------------

  < none :) >

-- 
Joel

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Simon Marchi]

>   * [<Simon + Andrew>] <PR backtrace/27147>
>     [GNU/Linux, sparc64] GDB is unable to print full stack trace (got "previous frame inner to this frame" errors)
>     https://sourceware.org/bugzilla/show_bug.cgi?id=27147
> 
>     As the subject says, it's an unwinder issue.  Simon was able
>     to reproduce, and tied this issue to a specific commit adding
>     multi-target support.
> 
>     Simon sent a patch to the PR which is a minimal-impact attempt
>     at fixing the issue, which referencing a a larger-scope attempt
>     at reducing our dependencies on "current_something" globals...

This is now pushed to both branches.

Simon

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Joel Brobecker]

> >   * [<Simon + Andrew>] <PR backtrace/27147>
> >     [GNU/Linux, sparc64] GDB is unable to print full stack trace (got "previous frame inner to this frame" errors)
> >     https://sourceware.org/bugzilla/show_bug.cgi?id=27147
> > 
> >     As the subject says, it's an unwinder issue.  Simon was able
> >     to reproduce, and tied this issue to a specific commit adding
> >     multi-target support.
> > 
> >     Simon sent a patch to the PR which is a minimal-impact attempt
> >     at fixing the issue, which referencing a a larger-scope attempt
> >     at reducing our dependencies on "current_something" globals...
> 
> This is now pushed to both branches.

Brilliant. Thanks Simon!

I believe this is it, for the list of known issues. I'll send another
dated update this weekend, and will propose we generate the 10.2 release
the following weekend (Mar 13-14).

-- 
Joel

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Keith Seitz]

On 3/4/21 10:28 PM, Joel Brobecker wrote:

> I believe this is it, for the list of known issues. I'll send another
> dated update this weekend, and will propose we generate the 10.2 release
> the following weekend (Mar 13-14).
> 

There is at least one critical bug that hasn't made it to gdb-10-branch:

commit e77b0004dd114d6ddf3bb92b521b2854341f3f85
Author: Tom de Vries <tdevries@suse.de>
Date:   Fri Feb 5 17:47:07 2021 +0100

    [gdb/symtab] Handle DW_TAG_type_unit in process_psymtab_comp_unit

This is currently generating many segfaults during testing (across
{unix,native-gdbserver,native-extended-gdbserver}/-m{32,64} on x86_64).

Is it too late to include this?

Keith

PS. My testing shows over *500* core files generated while testing the
aforementioned targets. I doubt this is the only problem.

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Joel Brobecker]

Hi Keith,

> On 3/4/21 10:28 PM, Joel Brobecker wrote:
> 
> > I believe this is it, for the list of known issues. I'll send another
> > dated update this weekend, and will propose we generate the 10.2 release
> > the following weekend (Mar 13-14).
> > 
> 
> There is at least one critical bug that hasn't made it to gdb-10-branch:
> 
> commit e77b0004dd114d6ddf3bb92b521b2854341f3f85
> Author: Tom de Vries <tdevries@suse.de>
> Date:   Fri Feb 5 17:47:07 2021 +0100
> 
>     [gdb/symtab] Handle DW_TAG_type_unit in process_psymtab_comp_unit
> 
> This is currently generating many segfaults during testing (across
> {unix,native-gdbserver,native-extended-gdbserver}/-m{32,64} on x86_64).
> 
> Is it too late to include this?

Not at all!

I reviewed the test, and it conveniently already has a PR number :-).
And looking at the patch, outside of causing a build failure, with
the fact that it is the last step before a call to abort, I think
we can confidently say that the patch can only be safer than the
alternative ;-).

So OK for me to backport.

Thanks for pointing it out, Keith!
-- 
Joel

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Kevin Buettner]

I know of two small commits which fix build problems on OSes using
glibc-2.33. These include Fedora 34 and Fedora Rawhide.  I think
these are low risk and I'm willing to apply them to gdb-10-branch:

See:

Fix aarch64-linux-hw-point.c build problem:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=665af52ec2a52184d39a76d6e724fa4733dbab3c

amd64-linux-siginfo.c: Adjust include order to avoid gnulib error:
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=8488c357ce4fc309d49c7b0224cf9574b68e8116

Okay?

Kevin

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Joel Brobecker]

Hi Kevin,

> I know of two small commits which fix build problems on OSes using
> glibc-2.33. These include Fedora 34 and Fedora Rawhide.  I think
> these are low risk and I'm willing to apply them to gdb-10-branch:
> 
> See:
> 
> Fix aarch64-linux-hw-point.c build problem:
> https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=665af52ec2a52184d39a76d6e724fa4733dbab3c
> 
> amd64-linux-siginfo.c: Adjust include order to avoid gnulib error:
> https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=8488c357ce4fc309d49c7b0224cf9574b68e8116
> 
> Okay?

OK for me; the only thing that's missing before backporting is
that we create a GDB PR for each of these changes that documents
the problem they are fixing, and add those PRs to the ChangeLog
entry and commit message. Would you mind doing that?

Thank you!
-- 
Joel

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Joel Brobecker]

> > commit e77b0004dd114d6ddf3bb92b521b2854341f3f85
> > Author: Tom de Vries <tdevries@suse.de>
> > Date:   Fri Feb 5 17:47:07 2021 +0100
> > 
> >     [gdb/symtab] Handle DW_TAG_type_unit in process_psymtab_comp_unit
> > 
> > This is currently generating many segfaults during testing (across
> > {unix,native-gdbserver,native-extended-gdbserver}/-m{32,64} on x86_64).
> > 
> > Is it too late to include this?
> 
> Not at all!
> 
> I reviewed the test, and it conveniently already has a PR number :-).
> And looking at the patch, outside of causing a build failure, with
> the fact that it is the last step before a call to abort, I think
> we can confidently say that the patch can only be safer than the
> alternative ;-).
> 
> So OK for me to backport.

Since the patch was all but ready to be backported, I took care of it.

-- 
Joel

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Kevin Buettner]

On Sat, 6 Mar 2021 10:06:29 +0400
Joel Brobecker <brobecker@adacore.com> wrote:

> Hi Kevin,
> 
> > I know of two small commits which fix build problems on OSes using
> > glibc-2.33. These include Fedora 34 and Fedora Rawhide.  I think
> > these are low risk and I'm willing to apply them to gdb-10-branch:
> > 
> > See:
> > 
> > Fix aarch64-linux-hw-point.c build problem:
> > https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=665af52ec2a52184d39a76d6e724fa4733dbab3c
> > 
> > amd64-linux-siginfo.c: Adjust include order to avoid gnulib error:
> > https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=8488c357ce4fc309d49c7b0224cf9574b68e8116
> > 
> > Okay?  
> 
> OK for me; the only thing that's missing before backporting is
> that we create a GDB PR for each of these changes that documents
> the problem they are fixing, and add those PRs to the ChangeLog
> entry and commit message. Would you mind doing that?

Yes, I'll do that.

Kevin

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Tom de Vries]

On 3/5/21 2:56 PM, Keith Seitz wrote:
> On 3/4/21 10:28 PM, Joel Brobecker wrote:
> 
>> I believe this is it, for the list of known issues. I'll send another
>> dated update this weekend, and will propose we generate the 10.2 release
>> the following weekend (Mar 13-14).
>>
> 
> There is at least one critical bug that hasn't made it to gdb-10-branch:
> 
> commit e77b0004dd114d6ddf3bb92b521b2854341f3f85
> Author: Tom de Vries <tdevries@suse.de>
> Date:   Fri Feb 5 17:47:07 2021 +0100
> 
>     [gdb/symtab] Handle DW_TAG_type_unit in process_psymtab_comp_unit
> 
> This is currently generating many segfaults during testing (across
> {unix,native-gdbserver,native-extended-gdbserver}/-m{32,64} on x86_64).
> 
> Is it too late to include this?
> 
> Keith
> 
> PS. My testing shows over *500* core files generated while testing the
> aforementioned targets. I doubt this is the only problem.
> 

Now that that commit has been backported, I ran the gdb testsuite on
gdb-10-branch with target board unix/gdb:debug_flags=-gdwarf-5 and ran into:
...
FAIL: gdb.fortran/function-calls.exp: p
derived_types_and_module_calls::pass_cart_nd(c_nd)
...
which should be fixed by backporting:
...
commit cf2b20752995e6f10d88afc49166e729c33beb48
Author: Tom de Vries <tdevries@suse.de>
Date:   Tue Feb 9 23:28:16 2021 +0100

    [gdb/symtab] Fix element type modification in read_array_type
...

So we could include that one as well.

Thanks,
- Tom

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Joel Brobecker]

> Now that that commit has been backported, I ran the gdb testsuite on
> gdb-10-branch with target board unix/gdb:debug_flags=-gdwarf-5 and ran into:
> ...
> FAIL: gdb.fortran/function-calls.exp: p
> derived_types_and_module_calls::pass_cart_nd(c_nd)
> ...
> which should be fixed by backporting:
> ...
> commit cf2b20752995e6f10d88afc49166e729c33beb48
> Author: Tom de Vries <tdevries@suse.de>
> Date:   Tue Feb 9 23:28:16 2021 +0100
> 
>     [gdb/symtab] Fix element type modification in read_array_type
> ...
> 
> So we could include that one as well.

Sounds good to me. Would you mind taking care of it?

And would you mind also changing the "Target Milestone" in
the PR database to 10.2 once done?

Thank you!
-- 
Joel

Re: GDB 10.2 release (respin) -- 2021-02-28 Update

[Tom de Vries]

On 3/7/21 6:54 AM, Joel Brobecker wrote:
>> Now that that commit has been backported, I ran the gdb testsuite on
>> gdb-10-branch with target board unix/gdb:debug_flags=-gdwarf-5 and ran into:
>> ...
>> FAIL: gdb.fortran/function-calls.exp: p
>> derived_types_and_module_calls::pass_cart_nd(c_nd)
>> ...
>> which should be fixed by backporting:
>> ...
>> commit cf2b20752995e6f10d88afc49166e729c33beb48
>> Author: Tom de Vries <tdevries@suse.de>
>> Date:   Tue Feb 9 23:28:16 2021 +0100
>>
>>     [gdb/symtab] Fix element type modification in read_array_type
>> ...
>>
>> So we could include that one as well.
> 
> Sounds good to me. Would you mind taking care of it?
> 
> And would you mind also changing the "Target Milestone" in
> the PR database to 10.2 once done?
> 
> Thank you!
> 

Done :)

Thanks,
- Tom