Re: master d8da850: Document cairo-related fallout to font-backend settings

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Stefan Kangas]

Robert Pluim <rpluim@gmail.com> writes:

> +Note also that 'FontBackend' settings in .Xdefaults or .Xresources, or
> +'font-backend' frame parameter settings in your init files, may need
> +to be adjusted, as 'xft' is no longer a valid backend when using
> +Cairo.  Use either 'ftcr' or 'ftcrhb' instead for FreeType and
> +FreeType + HarfBuzz respectively.  'x' is still a valid backend.

Should we say that FreeType + Harfbuzz is probably the correct choice
for most users, and perhaps also why?  My concern is that people might
default to the shorter option and we could see unnecessary bug reports
about incorrect composition, etc.

Best regards,
Stefan Kangas

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Mon, 20 Jan 2020 22:05:23 +0100, Stefan Kangas <stefan@marxist.se> said:

    Stefan> Robert Pluim <rpluim@gmail.com> writes:
    >> +Note also that 'FontBackend' settings in .Xdefaults or .Xresources, or
    >> +'font-backend' frame parameter settings in your init files, may need
    >> +to be adjusted, as 'xft' is no longer a valid backend when using
    >> +Cairo.  Use either 'ftcr' or 'ftcrhb' instead for FreeType and
    >> +FreeType + HarfBuzz respectively.  'x' is still a valid backend.

    Stefan> Should we say that FreeType + Harfbuzz is probably the correct choice
    Stefan> for most users, and perhaps also why?  My concern is that people might
    Stefan> default to the shorter option and we could see unnecessary bug reports
    Stefan> about incorrect composition, etc.

The correct choice depends on how their Emacs has been built; ftcr and
ftcrhb are mutually exclusive [1](see my earlier update to the elisp
manual). I guess we could say something like this:

diff --git a/etc/NEWS b/etc/NEWS
index 6cc392970f..55cf59283c 100644
--- a/etc/NEWS
+++ b/etc/NEWS
@@ -34,9 +34,9 @@ your font settings.
 Note also that 'FontBackend' settings in .Xdefaults or .Xresources, or
 'font-backend' frame parameter settings in your init files, may need
 to be adjusted, as 'xft' is no longer a valid backend when using
-Cairo.  Use either 'ftcr' or 'ftcrhb' instead for FreeType and
-FreeType + HarfBuzz respectively.  'x' is still a valid backend.
-
+Cairo.  Use 'ftcrhb' if your Emacs was built with HarfBuzz support,
+and 'ftcr' otherwise.  You can this determine by checking
+'system-configuration-features'.  'x' is still a valid backend.
 
 ---
 ** The ftx font backend driver has been removed.

Footnotes:
[1]  My master plan to remove as many different variations as possible
     seems to get shot down every time I suggest it :-)

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Stefan Kangas]

Robert Pluim <rpluim@gmail.com> writes:

> The correct choice depends on how their Emacs has been built; ftcr and
> ftcrhb are mutually exclusive [1](see my earlier update to the elisp
> manual). I guess we could say something like this:

Indeed, I read that after I wrote the above.

> -Cairo.  Use either 'ftcr' or 'ftcrhb' instead for FreeType and
> -FreeType + HarfBuzz respectively.  'x' is still a valid backend.
> -
> +Cairo.  Use 'ftcrhb' if your Emacs was built with HarfBuzz support,
> +and 'ftcr' otherwise.  You can this determine by checking
> +'system-configuration-features'.  'x' is still a valid backend.

Thank you, that is much better.  Please install.

> [1]  My master plan to remove as many different variations as possible
>      seems to get shot down every time I suggest it :-)

BTW, if 'ftcr' only works when built without HarfBuzz, and 'ftcrhb'
only works with HarfBuzz, can't we just use 'ftcr' and detect HarfBuzz
automatically for the user?

Best regards,
Stefan Kangas

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Tue, 21 Jan 2020 00:12:22 +0100, Stefan Kangas <stefan@marxist.se> said:

    Stefan> Thank you, that is much better.  Please install.

Done. 
    >> [1]  My master plan to remove as many different variations as possible
    >> seems to get shot down every time I suggest it :-)

    Stefan> BTW, if 'ftcr' only works when built without HarfBuzz, and 'ftcrhb'
    Stefan> only works with HarfBuzz, can't we just use 'ftcr' and detect HarfBuzz
    Stefan> automatically for the user?

I see two options:

1. give the Qftcrhb symbol the same name as Qftcr, which might do the
right thing, but would make it harder to work out whether HarfBuzz is
actually being used.

2. just say "If youʼre going to be modern, be modern: Cairo requires
HarfBuzz" (note that this is the path that Pango is taking).

Robert

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Date: Tue, 21 Jan 2020 08:44:39 +0100
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
> I see two options:
> 
> 1. give the Qftcrhb symbol the same name as Qftcr, which might do the
> right thing, but would make it harder to work out whether HarfBuzz is
> actually being used.
> 
> 2. just say "If youʼre going to be modern, be modern: Cairo requires
> HarfBuzz" (note that this is the path that Pango is taking).

I think 2 is the best alternative right now.  I don't think we should
make more radical decisions based on such incomplete experience, both
with the Cairo build and with HarfBuzz.

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Tue, 21 Jan 2020 17:32:23 +0200, Eli Zaretskii <eliz@gnu.org> said:

    >> From: Robert Pluim <rpluim@gmail.com>
    >> Date: Tue, 21 Jan 2020 08:44:39 +0100
    >> Cc: Emacs developers <emacs-devel@gnu.org>
    >> 
    >> I see two options:
    >> 
    >> 1. give the Qftcrhb symbol the same name as Qftcr, which might do the
    >> right thing, but would make it harder to work out whether HarfBuzz is
    >> actually being used.
    >> 
    >> 2. just say "If youʼre going to be modern, be modern: Cairo requires
    >> HarfBuzz" (note that this is the path that Pango is taking).

    Eli> I think 2 is the best alternative right now.  I don't think we should
    Eli> make more radical decisions based on such incomplete experience, both
    Eli> with the Cairo build and with HarfBuzz.

So a configure.ac patch that says 'if cairo && ! harfbuzz error_out'
is pre-approved? Or did you have something else in mind?

Robert

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Cc: stefan@marxist.se,  emacs-devel@gnu.org
> Date: Tue, 21 Jan 2020 16:39:53 +0100
> 
>     >> I see two options:
>     >> 
>     >> 1. give the Qftcrhb symbol the same name as Qftcr, which might do the
>     >> right thing, but would make it harder to work out whether HarfBuzz is
>     >> actually being used.
>     >> 
>     >> 2. just say "If youʼre going to be modern, be modern: Cairo requires
>     >> HarfBuzz" (note that this is the path that Pango is taking).
> 
>     Eli> I think 2 is the best alternative right now.  I don't think we should
>     Eli> make more radical decisions based on such incomplete experience, both
>     Eli> with the Cairo build and with HarfBuzz.
> 
> So a configure.ac patch that says 'if cairo && ! harfbuzz error_out'
> is pre-approved? Or did you have something else in mind?

No, I meant to say this in the docs.  Sorry for being unclear.  IOW,
recommend that people build with Cairo and HarfBuzz as the main
configuration.

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Tue, 21 Jan 2020 17:47:21 +0200, Eli Zaretskii <eliz@gnu.org> said:
    >> So a configure.ac patch that says 'if cairo && ! harfbuzz error_out'
    >> is pre-approved? Or did you have something else in mind?

    Eli> No, I meant to say this in the docs.  Sorry for being unclear.  IOW,
    Eli> recommend that people build with Cairo and HarfBuzz as the main
    Eli> configuration.

Iʼve put some stuff in NEWS to that effect. Would you be ok with
configure warning about Xft and about Cairo without HarfBuzz?

Robert

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Cc: stefan@marxist.se,  emacs-devel@gnu.org
> Date: Wed, 22 Jan 2020 08:22:29 +0100
> 
> Would you be ok with configure warning about Xft and about Cairo
> without HarfBuzz?

Can you show the proposed text of the warning?

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Wed, 22 Jan 2020 17:52:49 +0200, Eli Zaretskii <eliz@gnu.org> said:

    >> From: Robert Pluim <rpluim@gmail.com>
    >> Cc: stefan@marxist.se,  emacs-devel@gnu.org
    >> Date: Wed, 22 Jan 2020 08:22:29 +0100
    >> 
    >> Would you be ok with configure warning about Xft and about Cairo
    >> without HarfBuzz?

    Eli> Can you show the proposed text of the warning?

if test "$with_xft" = yes; then
AC_MSG_WARN([This configuration uses libXft,
which has a number of font rendering issues, and is being considered for
removal in the next release of Emacs.  Please consider using Cairo + HarfBuzz
instead (they are auto-detected if the relevant development headers are
installed).])
fi

if test "${HAVE_CAIRO}" = "yes" && test "${HAVE_HARFBUZZ}" = no; then
   AC_MSG_WARN([We recommend the use of HarfBuzz when using Cairo,
   please install HarfBuzz development packages])
fi

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Cc: stefan@marxist.se,  emacs-devel@gnu.org
> Date: Wed, 22 Jan 2020 17:11:30 +0100
> 
>     Eli> Can you show the proposed text of the warning?
> 
> if test "$with_xft" = yes; then
> AC_MSG_WARN([This configuration uses libXft,
> which has a number of font rendering issues, and is being considered for
> removal in the next release of Emacs.  Please consider using Cairo + HarfBuzz
> instead (they are auto-detected if the relevant development headers are
> installed).])
> fi
> 
> if test "${HAVE_CAIRO}" = "yes" && test "${HAVE_HARFBUZZ}" = no; then
>    AC_MSG_WARN([We recommend the use of HarfBuzz when using Cairo,
>    please install HarfBuzz development packages])
> fi

Sounds good, thanks.

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Date: Wed, 22 Jan 2020 08:22:29 +0100
> Cc: stefan@marxist.se, emacs-devel@gnu.org
> 
> >>>>> On Tue, 21 Jan 2020 17:47:21 +0200, Eli Zaretskii <eliz@gnu.org> said:
>     >> So a configure.ac patch that says 'if cairo && ! harfbuzz error_out'
>     >> is pre-approved? Or did you have something else in mind?
> 
>     Eli> No, I meant to say this in the docs.  Sorry for being unclear.  IOW,
>     Eli> recommend that people build with Cairo and HarfBuzz as the main
>     Eli> configuration.
> 
> Iʼve put some stuff in NEWS to that effect.

It looks like you did that on master.  Was that intentional?

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Wed, 22 Jan 2020 18:53:46 +0200, Eli Zaretskii <eliz@gnu.org> said:

    >> From: Robert Pluim <rpluim@gmail.com>
    >> Date: Wed, 22 Jan 2020 08:22:29 +0100
    >> Cc: stefan@marxist.se, emacs-devel@gnu.org
    >> 
    >> >>>>> On Tue, 21 Jan 2020 17:47:21 +0200, Eli Zaretskii <eliz@gnu.org> said:
    >> >> So a configure.ac patch that says 'if cairo && ! harfbuzz error_out'
    >> >> is pre-approved? Or did you have something else in mind?
    >> 
    Eli> No, I meant to say this in the docs.  Sorry for being unclear.  IOW,
    Eli> recommend that people build with Cairo and HarfBuzz as the main
    Eli> configuration.
    >> 
    >> Iʼve put some stuff in NEWS to that effect.

    Eli> It looks like you did that on master.  Was that intentional?

Yes. Did you want it on emacs-27? Cairo is available but not enabled
by default there.

Robert

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Cc: stefan@marxist.se,  emacs-devel@gnu.org
> Date: Wed, 22 Jan 2020 17:59:17 +0100
> 
>     >> Iʼve put some stuff in NEWS to that effect.
> 
>     Eli> It looks like you did that on master.  Was that intentional?
> 
> Yes. Did you want it on emacs-27? Cairo is available but not enabled
> by default there.

Ah, okay then.  I guess I succeeded to confuse myself.

Re: master d8da850: Document cairo-related fallout to font-backend settings

[T.V Raman]

Stefan Kangas <stefan@marxist.se> writes:

Also, perhaps update frames.texi to indicate that ftcr and ftcrhb are
mutually incompatible. Having both doesn't appear to hurt right now, but
 might perhaps cause issues later?> Robert Pluim <rpluim@gmail.com> writes:
>
>> The correct choice depends on how their Emacs has been built; ftcr and
>> ftcrhb are mutually exclusive [1](see my earlier update to the elisp
>> manual). I guess we could say something like this:
>
> Indeed, I read that after I wrote the above.
>
>> -Cairo.  Use either 'ftcr' or 'ftcrhb' instead for FreeType and
>> -FreeType + HarfBuzz respectively.  'x' is still a valid backend.
>> -
>> +Cairo.  Use 'ftcrhb' if your Emacs was built with HarfBuzz support,
>> +and 'ftcr' otherwise.  You can this determine by checking
>> +'system-configuration-features'.  'x' is still a valid backend.
>
> Thank you, that is much better.  Please install.
>
>> [1]  My master plan to remove as many different variations as possible
>>      seems to get shot down every time I suggest it :-)
>
> BTW, if 'ftcr' only works when built without HarfBuzz, and 'ftcrhb'
> only works with HarfBuzz, can't we just use 'ftcr' and detect HarfBuzz
> automatically for the user?
>
> Best regards,
> Stefan Kangas
>

-- 

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Tue, 21 Jan 2020 06:54:36 -0800, "T.V Raman" <raman@google.com> said:

    TVR> Stefan Kangas <stefan@marxist.se> writes:
    TVR> Also, perhaps update frames.texi to indicate that ftcr and ftcrhb are
    TVR> mutually incompatible. Having both doesn't appear to hurt right now, but
    TVR>  might perhaps cause issues later?

I updated frames.texi on emacs-27 earlier in the week along those
lines. Having both doesnʼt hurt, they're not incompatible, you just can
only have one or the other (and you canʼt switch between them dynamically).

Robert

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Robert Pluim]

>>>>> On Tue, 21 Jan 2020 16:38:01 +0100, Robert Pluim <rpluim@gmail.com> said:

>>>>> On Tue, 21 Jan 2020 06:54:36 -0800, "T.V Raman" <raman@google.com> said:
    TVR> Stefan Kangas <stefan@marxist.se> writes:
    TVR> Also, perhaps update frames.texi to indicate that ftcr and ftcrhb are
    TVR> mutually incompatible. Having both doesn't appear to hurt right now, but
    TVR> might perhaps cause issues later?

    Robert> I updated frames.texi on emacs-27 earlier in the week along those
    Robert> lines. Having both doesnʼt hurt, they're not incompatible, you just can
    Robert> only have one or the other (and you canʼt switch between them dynamically).

Let me qualify that: you can in fact use the ftcr font backend even if
your emacs is built with HarfBuzz, but you need to make that decision
at frame creation time. Iʼll reword NEWS and frames.texi to that
effect.

Robert

Re: master d8da850: Document cairo-related fallout to font-backend settings

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Date: Mon, 20 Jan 2020 22:50:27 +0100
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
> +Cairo.  Use 'ftcrhb' if your Emacs was built with HarfBuzz support,
> +and 'ftcr' otherwise.  You can this determine by checking
> +'system-configuration-features'.  'x' is still a valid backend.

LGTM, except that "You can this determine" needs to switch the order
of words.

Thanks.