Re: GDB 15.1 release update (2024-03-23)

[Thiago Jung Bauermann <thiago.bauermann@linaro.org>]

Hello Joel,

Joel Brobecker <brobecker@adacore.com> writes:

> First, a reminder of our target dates:
>
>   - Pre-release:    13-14 April 2024
>   - Release (15.1): 27-28 April 2024
>
> As we'll see in the recap below, things are moving forward, albeit
> slowly, and we also have one (and maybe two) new issues which have
> been reported. As a result, I'm starting to think that maybe
> we won't make the dates above, which isn't a complete surprise
> since we know those dates were a little on the optimistic side.
> We'll see!
>
> Meanwhile, below is a quick overview of where we stand in terms
> of known issues. Thank you to everyone for creating PRs, as
> this helps tremendously with tracking!
>
> Please do let us know if there are other issues you think should be
> fixed in GDB 15.

It's not an issue per se, but if possible I would like to get some
improvements for aarch64 MOPS (Memory Operations) instructions in
GDB 15.

These implement memcpy, memset and memmove with a sequence of
instructions that need to be executed in sequence, and thus can cause
issues when single-stepping through them, and also regarding detection
of when they trigger a watchpoint.

I'm striving to post patches in the next two to three weeks.

Tom de Vries recently opened a bug about adding a testcase for these
instructions:

[gdb/testsuite] Add memcpy, memmove, memset watchpoint tests
https://sourceware.org/bugzilla/show_bug.cgi?id=31484

> Other Ongoing Items:
> --------------------
>
>   * [CarlEL] testsuite/31312
>     attach-many-short-lived-threads gives inconsistent results (POWER 10)
>     https://sourceware.org/bugzilla/show_bug.cgi?id=31312
>
>         Looked like the discussions have reached a consensus of accepting
>         this failure (2024-03-09), but there was a second questin that
>         wasn't wasn't understood yet, so Carl tried to dig deeper.
>         Meanwhile, Thiago indicated he's seeing the same kind of behavior
>         on AArch64 Linux (sometimes), and now has some leads he is
>         pursuing (go Thiago!)

Thank you! :-)

>         Thiago posted a patch series that fixes the problem, but
>         with one known regression he hasn't had time to investigate:
>         https://sourceware.org/pipermail/gdb-patches/2024-March/207467.html

Pedro suggested an alternative approach to fix the problem. I tried that
suggestion (or something similar to it) before posting the patch series
and there was some other problem, but I can't recall the details. I'll
try again next week and see how it goes.

--
Thiago