SKIA on Windows Roadmap (ESC agenda item)

Telesto telesto at surfxs.nl
Thu Apr 23 19:16:53 UTC 2020


Hi all,

Skia should be topic/ agenda item for ESC team. There needs be some kind 
of roadmap/ timeline/ vision:

  * Skia the default for 7.0 release on Windows?

* What about the future GDI (incl. Hardware Acceleration)/OpenGL Windows 
backend? Should those be phased out? If so, what's the intended 
time-table at this point in time:

- Active support of the other backend (bug fixing)

- Passively supporting the other backends; so being available?
    ** When do we move the setting to Expert configuration
    ** When should the old code be removed

  * What is future of Skia for Linux/Mac? What does this mean for other 
backends? Cairo/ X11...
    - A number of the base QA members use Linux. Is there backend 
comparable to Skia available; using 'the same' code path?

* QA concerns (based on: bug 132323)
- Firebird experience: flood of bugs after removing from experimental

- Harfbuzz experience: quite a number number of performances issues at 
the beginning

- One developer for Skia/ Single point of failure?


Personal vision:

* Active support for multiple backends are hard to maintain at QA and 
DEV level, so less is more. Backend options should be 'invisible' as 
fast as reasonable possible. Only objection: different backends have 
different code paths; can be an alternative. Example: Skia XOR issue 
(affecting Raster & Vulkan); without GDI no way out.

* Skia looks promising; not seeing major problems. They should be rather 
obvious to see. Everybody uses the UI.  So no Firebird experience

* Passive support should not be GUI advertised for to long. Bugs tend to 
sneak in, while everybody else is using the default. So settings for 
backends should move Expert configuration pretty fast. It should be 
decided at ESC level; it's not UI-only issue. It's also affecting QA & DEV

* Removing code. Should be hold off for a while. Second next major 
release. The old layout engine was removed a little to soon for my 
taste. Can be useful as fall-back.

* Prevent a dependence on a single developer; there needs to be a 
'backup' developer who wants to step in, if needed. What about 
documentation?

* Developer available to squash the initial bunch of Skia related bugs 
after release (to assure the QA department) . So everything is 'back to 
normal' at 7.0.2 . If Skia contains bugs at 7.0 so be it. LibreOffice 
fresh can contain bugs. Firebird lacked a 'testerbase' while under 
development and support afterwards didn't win awards either.

* OpenGL. Drop active support with 7.0. Moving to expert configuration 
with LibreOffice 7.1. Code removal at 7.3. Deprecation should be 
mentioned in the Release Notes

* GDI: Active support dropped at 7.2. Passive support until 7.4. Moving 
to expert configuration at 8.0. Mention this roadmap in the Release Notes

Regards,
Telesto




More information about the LibreOffice mailing list