bug#38442: 27.0.50; segmentation fault switching to cairo

[Juanma Barranquero <lekktu@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1098 bytes --]

On Mon, Dec 2, 2019 at 6:48 PM Eli Zaretskii <eliz@gnu.org> wrote:

> The uniscribe backend will not be used unless Emacs is started with it
> being in the list of backends to register.

That's what I thought.

> Every font that can use Uniscribe can also use HarfBuzz.  So this is
> not a problem, and users have no reason to wish to use Uniscribe when
> they restore their sessions.

That's an argument to remove support for Uniscribe. I imagine that, if we
haven't done that, it's likely because there could be some problematic
font that would behave better with Uniscribe than Harfbuzz, isn't it?

> Moreover, unless Emacs is started with Uniscribe (via the -xrm
> command-line switch, or via the Registry), it will be unable to use
> Uniscribe even if we tell it via the frameset.

Again, that was my understanding.

All in all, I think I would prefer to keep font-backend and filter it out
on restoring when/if necessary; but I don't have a compelling use case,
so I'll just filter it out entirely. We can always change it later if the
need arises.

I'll commit it ASAP.

Thanks,

    Juanma

[-- Attachment #2: Type: text/html, Size: 1528 bytes --]