Re: scratch/igc: Keeping a TODO list?

[Pip Cet]

"Eli Zaretskii" <eliz@gnu.org> writes:

>> Date: Sun, 12 Jan 2025 15:50:04 +0000
>> Cc: Stefan Kangas <stefankangas@gmail.com>, emacs-devel@gnu.org, gerd@gnu.org
>> From:  Pip Cet via "Emacs development discussions." <emacs-devel@gnu.org>
>>
>> 8. use malloc()/memcpy()/free() rather than realloc(), remove the arena
>> parking when a root is "resized"
>
> This will be a performance hit, since in many cases realloc doesn't
> move data at all.  So I don't think it's a good idea.

(I meant only for igc roots, of course.  Realloc for buffer text is
perfectly okay.)

If memcpy()ing an unprotected root is a performance hit, the root is
way, way too large.

>> 10. Ban WIDE_EMACS_INT + HAVE_MPS builds again.
>
> I won't agree to this change any time soon, sorry.

I'll move those two off the TODO list, then.

I didn't put "reduce the number of roots created by gtkutil.c" on either
list.  I've established gtkutil.c currently leaks memory in all builds,
and I'd like to fix that before deciding how many roots are acceptable.

My understanding is Po Lu is the relevant person to include, but I think
it would be the polite thing to clearly document the memory leaks on the
master branch first rather than including him in a lengthy discussion he
might not be interested in.

My guess is we should move to a ref-counted structure for
glib_user_data, which would reduce the number of MPS roots.

>> Maybe it's more interesting what's not on the list:
>
> I think we should have a TODO list of the necessary jobs that are
> prerequisites for landing the branch.  Developments that are not
> necessary for that are best kept separate, since they can wait until
> after we land the branch.

Agreed.

Pip