[cairo] Fwd: Re: Time for a release ?

[Bryce Harrington]

On Tue, Nov 17, 2020 at 10:14:29PM +0100, Heiko Lewin wrote:
> I would say "Feel free top open a bug report", but is this related to
> https://gitlab.freedesktop.org/cairo/cairo/-/merge_requests/1 ?
> This could depend on the config of freetype, or cairos default behviour
> could have changed.
> I don't know - a small test case showing the problem would be nice.
> 
> 
> On 11/17/20 7:38 PM, Dave Yeo wrote:
> > On 11/17/20 06:10 AM, Uli Schlachter wrote:
> > > > [...] however I cannot remember the test-suite passing at
> > > > any time,
> > > Yup, I never saw the test suite passing either.
> > 
> > My understanding is that the testsuite requires exactly the right
> > versions of the various fonts, ghostscript and likely other things and
> > likely only ever passed on the original developers setup.

There's also a strong coupling to pixman rendering behavior.  From my
own archaeology into the testsuite it seems after pixman was split out
from Cairo things got more complicated since if you have a different
version of pixman than the version used by whomever last regenerated the
reference images, your test output likely won't match.

Given this (and the font dependence), I found the testsuite not to be
helpful when producing releases.  For CI, YMMV but I wonder if it might
end up being more maintenance work than worth.

However, the testsuite is still useful for local regression testing.
Basically, regenerate the reference images against master on your own
system locally, and then run the testsuite on your branch to spot
rendering changes.

It's also useful if you're looking for legitimate rendering bugs, but
this requires manual eyeball checking of the output.  I caught some
issues this way doing this exercise a few years back, but there were
some antialiasing issues I didn't figure out.  There are also still
quite a few backend-specific rendering issues, and version of
ghostscript, et al can impact the results.

Bryce

> > In the real world, I tried using the latest (well recent) Cairo with
> > Firefox and I see some fonts where the beginning of the font is faint
> > or non-existing. An F where the vertical stroke is missing for
> > example, only at the beginning of lines and mostly where Firefox fakes
> > system menus. This might be partially caused by the system font but
> > still seems to be a regression.
> > Dave
> -- 
> cairo mailing list
> cairo at cairographics.org
> https://lists.cairographics.org/mailman/listinfo/cairo