Re: emacs-29 b8d2ec920f: Revert "Improve last change to xfaces.c" (05ece1eb8b)

[Po Lu]

Gregory Heytings <gregory@heytings.org> writes:

> It is telling that you "see no purpose" in discussing a change to code
> that was agreed upon after 300 posts in a bug report.

Sorry, but I don't see where in the discussion the use of a bitmask as a
variable was actually discussed.  My change does not touch any of the
code which was discussed.

> That variable is of no interest whatsoever to our users.  It is there
> only for debugging purposes, and is once again only meant to be used
> by the few users who understand subtle technicalities in the face
> realization code.

Then the best course of action is to just remove the variable.

> Aha.  That's your understanding of a "discussion", then: you say
> something, and act immediately, without waiting for a potential
> answer. And then claim that there was a "discussion" because you said
> something. Oh, in fact no, that's not even what happened: you reverted
> before saying anything.

Only because you reverted first.

> You would not ask that question if you had read the explanation in
> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=59347#331.  And the fact
> you ask that question shows that you did not read it.

Anyway, AFAIK :font can be a font-entity and not a font-spec, in which
case it is definitely not okay to touch :extra under the Haiku font
backend: it contains two indices into the system-wide font and family
arrays which should not be changed.

> The relevant parts of that enum have not changed at all since they
> were introduced fifteen years ago.

And will you guarantee that will always be the case?

> Nobody should do that.  The docstring clearly said: "There is no
> reason to change that value except for debugging purposes."

Alas, what ``should'' be is very different from reality.  Someone will
set the unused bits to random garbage, and someone elses nice plan for
adding extra fields to a font object will be broken by that change.

> Oh yes, I see.  FONT_SPEC_MAX is (and has always been) 15, but
> clearly, for a reason that hasn't happened in the past fifteen years,
> "at some point in the future", it will become 30, and that will be
> problematic on 32-bit computers.  And you tell me that what I write is
> "nonsense".

It could change, and once it does I can almost guarantee that nobody
will think to update the ``obscure'' bitmask exposed to Lisp that users
``are not supposed'' to set.

> Anyway, I don't want to deal with this anymore.  I probably spent
> about 50 hours on that bug, that's more than enough.

Fair enough.