On Fri, Aug 2, 2019 at 10:21 AM Uli S= chlachter <psychon@znc.in> wrot= e:

Hi,

On 01.08.19 23:00, Matthias Clasen wrote:
> But I'd still like to understand what the rationale is for a) this= being
> separate from hint metrics
> and b) this not being settable.

the rationale basically is "the test suite's reference images are<= br> correct by definition, so how can we make cairo generate this output"?=
According to "git blame", I introduced this option back when I st= ill
tried to sort out the mess that is the test suite:

commit cee1dadc66107e1cc6bcac7718e0b67969141876
Author: Uli Schlachter <psychon@znc.in>
Date:=C2=A0 =C2=A0Thu Oct 21 22:14:57 2010 +0200

=C2=A0 =C2=A0 Other backends: Set round_glyph_positions to OFF

=C2=A0 =C2=A0 This sets CAIRO_ROUND_GLYPH_POS_OFF in all surface backends t= hat
didn't use
=C2=A0 =C2=A0 CAIRO_ROUND_GLYPH_POS_ON.

=C2=A0 =C2=A0 Signed-off-by: Uli Schlachter <psychon@znc.in>

commit d24e9881de7b392cfce1d82c9861553ece558aaf
Author: Uli Schlachter <psychon@znc.in>
Date:=C2=A0 =C2=A0Thu Oct 21 14:03:38 2010 +0200

=C2=A0 =C2=A0 raster backends: Set round_glpyh_positions to ON

=C2=A0 =C2=A0 This fixes "text-rotate" in the test suite for the = image backend and
=C2=A0 =C2=A0 "overlapping-glyphs" for the xcb backend.

=C2=A0 =C2=A0 Signed-off-by: Uli Schlachter <psychon@znc.in>

commit b514863a0ee8e6c4f0a994f6f5e7db18ce290e79
Author: Uli Schlachter <psychon@znc.in>
Date:=C2=A0 =C2=A0Thu Oct 21 13:59:25 2010 +0200

=C2=A0 =C2=A0 Actually implement round_glpyh_positions

=C2=A0 =C2=A0 The previous commit only added this option and made sure it g= ets
set, but it
=C2=A0 =C2=A0 didn't actually have any effect. This commit now implemen= ts this option.

=C2=A0 =C2=A0 Signed-off-by: Uli Schlachter <psychon@znc.in>

commit 6bfe71124b56b496056b77b3b51eef4d656ccf54
Author: Uli Schlachter <psychon@znc.in>
Date:=C2=A0 =C2=A0Thu Oct 21 13:52:33 2010 +0200

=C2=A0 =C2=A0 font options: Add private round_glpyh_positions field

=C2=A0 =C2=A0 Signed-off-by: Uli Schlachter <psychon@znc.in>

Looking at the git history for the reference images... I am not quite
sure where it came from. Back in 2010, I was assuming that the reference images are correct, I guess. Since then, I saw others fixing test
failures by just declaring the current cairo output as the reference. I
never tried to do that since I do not what each test actually really
tests and if the difference that makes the test fail is a bug or not.

What to do with this information... well, I do not know. If you want,
you can revert the above four commits, but I would hope that this does
not regress the test suite (=3D some higher power has to declare that the current reference images are wrong and the new output is correct and
should be the new reference image).

I might suggest to simply reuse hi= nt-metrics for this purpose, based

on the theory that if you don&= #39;t want hinted metrics, you probably don't

want rounded gl= yph positions either.