GDB 15.1 release update (2024-05-19)

GDB 15.1 release update (2024-05-19)

[Joel Brobecker]

Hi everyone,

We're getting very close. In fact, I think I'd like to propose that
we branch and pre-release next weekend (May 25-26). As you'll see
below, the recap shows that we're not 100 clean, but I still think
we should just go ahead, here's my reasoning:

  - gdb/31312 (attach-many-short-lived-threads gives inconsistent
    results)

        This one has all the markers of an issue that'll be difficult
        to track down and fix, while at the same time the conditions
        during which it happens are also fairly unique, and IIUC,
        this is only on PowerPC. So the impact IMO is low.

  - gdb/31751 (Change dwarf_synchronous back to true)

        This one is a trivial patch which I expect will get in soon.

  - The other two are not blockers.

This is why I propose next weekend for branching and pre-release.
And if we manage to fix the various issues we've identified, we can
still look at backporting them to the gdb-15-branch.

Please let us know if you think we should wait. Otherwise, I will
do the branching and create the first pre-release for early testing.

Thank you!

Now, the recap:

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

  * [<unassigned>] symtab/30520
    Slow lookup_symbol_in_objfile
    https://sourceware.org/bugzilla/show_bug.cgi?id=30520

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

  * [Thiago] gdb/31312
    attach-many-short-lived-threads gives inconsistent results
    https://sourceware.org/bugzilla/show_bug.cgi?id=31312

        Reopened after Carl reported random errors on 2024-050-14.
        Looks like it might only happen quite rarely.

  * [TomT] gdb/31751
    Change dwarf_synchronous back to true
    https://sourceware.org/bugzilla/show_bug.cgi?id=31751

        For 15.1, we will simply re-enable synchronous DWARF reading:
        https://sourceware.org/pipermail/gdb-patches/2024-May/209267.html

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

  < none :) >

Not Blocking, But Keep An Eye On:
---------------------------------

  * [Thiago]
    Recent change in gdb.base/relativedebug.exp disables it in systems with statically linked libc
    https://sourceware.org/bugzilla/show_bug.cgi?id=31700

        A bit of a regression, but testsuite-only, so not super
        critical. On the other hand, a patch was posted on 2024-04-29,
        so hoping we can fix it relatively quickly. Worse case
        scenario, we push it after branching.

  * [Thiago]
    various improvements to aarch64 MOPS (Memory Operations)
    https://sourceware.org/bugzilla/show_bug.cgi?id=31666

    v3 patch series (2024-05-10):
    https://sourceware.org/pipermail/gdb-patches/2024-May/208822.html

-- 
Joel

Re: GDB 15.1 release update (2024-05-19)

[Tom Tromey]

>>>>> "Joel" == Joel Brobecker <brobecker@adacore.com> writes:

Joel>   - gdb/31312 (attach-many-short-lived-threads gives inconsistent
Joel>     results)

Joel>         This one has all the markers of an issue that'll be difficult
Joel>         to track down and fix, while at the same time the conditions
Joel>         during which it happens are also fairly unique, and IIUC,
Joel>         this is only on PowerPC. So the impact IMO is low.

Thiago pointed out that the failures that are seen now seem to be from a
different issue than the one originally filed in this bug, and Tom de
Vries agreed and closed this.

Joel>   - gdb/31751 (Change dwarf_synchronous back to true)
Joel>         This one is a trivial patch which I expect will get in soon.

I checked this in today.

Joel>   * [Thiago]
Joel>     Recent change in gdb.base/relativedebug.exp disables it in systems with statically linked libc
Joel>     https://sourceware.org/bugzilla/show_bug.cgi?id=31700

I sent an approval for this.

Tom

Re: GDB 15.1 release update (2024-05-19)

[Tom de Vries]

On 5/22/24 18:50, Tom Tromey wrote:
> Joel>   - gdb/31312 (attach-many-short-lived-threads gives inconsistent
> Joel>     results)
> 
> Joel>         This one has all the markers of an issue that'll be difficult
> Joel>         to track down and fix, while at the same time the conditions
> Joel>         during which it happens are also fairly unique, and IIUC,
> Joel>         this is only on PowerPC. So the impact IMO is low.
> 
> Thiago pointed out that the failures that are seen now seem to be from a
> different issue than the one originally filed in this bug, and Tom de
> Vries agreed and closed this.

Indeed at that point there was no evidence that the PR reoccurred, but 
now there may be.  I've filed this PR ( 
https://sourceware.org/bugzilla/show_bug.cgi?id=31832 ) with the 
details.  We can mark it a duplicate and reopen the original PR once 
that's confirmed.

Thanks,
- Tom