Hey everyone,
Event: https://inkscape.org/cals/event/1/
Attending: Charlotte, Martin, Sean Burke, Mike Terry, Tav
Next Meeting: June 30th at the later time:
https://inkscape.org/cals/event/64/
Martin talked about his rendering engine work, most of the tests are
passing. Remaining issues are fixing filter integration, rasters and
the conversion from INT32 formats and a couple of other issues. At the
end of the meeting Martin gave a demonstration of the test generation
python script and some of the ways this helps with the development and
testing.
Charlotte asked about the performance impact of converting images from
INT to FLOAT before being able to render it. This would be easier on
the GPU as it's always floating point, but we basically have no choice
since cairo can't paint INT to FLOAT and visa-versa so the stack has to
keep things consistant.
Tav and Martin talked about type conversions and the problem of
rendering at 2x scale to reduce artefacts and how that would impact
performance. Using the preview to do a better job at rendering for
example.
Tav has been struggling with how to handle variable font axis opsz
(optical size). A newer way of doing size font hinting in a smarter
way. We have a font instance at 512x scale and then scale everything
ourselves to get the right sizes. But this doesn't work for optical
size axis since it changes the shape of the glyph based on how large it
is on screen.
Martin talks about how to design the API with what we have with make
font instances capable and not break a bunch of things. A FontSet
object is planned to track font selections more generically.
Charlotte discovered that latex won't include pdfs generated by
capypdf. Generating reproducable version and will make an issue for it.
Roland talks about the Bug Accelerator Program; priority on more
serious issues. There's some follow up with some serious developers for
some of the issues. Reviewing the 1.4.x issues for moving to 1.5. This
week they're removing low priority issues to clear the decks for seeing
what the major blockers are.
Martin will be making an assessment in a week's time for the extension
to the bug accelerator project.
Sean reports that there is a crash while it builds it's cache, and it
might not be resolvable without refactoring. He recommend that for now
the new font dialog should be disabled. Martin adds that the new object
properties dialog should also be disabled if the large speed impacts
can't be fixed.
See you all at the next meeting, June 30th.
Best regards, Martin Owens
Hey Everyone,
Event: https://inkscape.org/cals/event/64/
Attending: Mc, Martin, René, pbs, Vaibhav
Next: Monday 22nd June; https://inkscape.org/cals/event/1/
Marc managed to get access to the windows store and he will upload
Inkscape 1.4.4 to the store as soon as he is free.
René has been preparing the latest updates. Mason 1.11.x is the highest
we can go, Mason 1.12 and above will require a python version which is
not available for the macOS versions that we support. Has been
investigating how to integrate all the deps, turbo jpeg is using
graphics magick instead of image magick 6.
PBS reports that capypdf is now available for his android CI which is
useful. We want capypdf available for all our platforms and we can't
make it available on macOS because of meson, and the C++23 requirements
for the PPA and AppImage.
We talk about graphics magic instead of image magic and the instability
of upgrading Gtk which caused knock on effects. Martin says the icc
jpeg profile function could be patched out if we wanted to restrict the
version.
Marc, René and pbs talk about build systems, windows, macOS and other
platforms. Possibly using Nix, or Conan instead which already has a lot
of packages available. But questions of it being invasive, although it
might make sense for package building and CI.
The only part of inkscape that requires image magic at all is a set of
internal bitmap extensions; most of them could be either ifdefed out or
replaced with internal functions.
Philip would like to package caapypdf as a 3rdparty submodule to fix
the C++23 and meson requirements which are easy to patch out and makes
building for intel mac and PPA much harder. Martin asks that an issue
be created upstream to let Jussi know about what we're action we're
taking down stream and why.
Philip asked Marc for PPA administration access. Marc says ok.
The Gtk and Gtkmm versions are always out of sync on debian systems and
this causes us problems. We talked about peel some more for the future
upgrade path.
Martin has been working on outputting images in the command line, for
no real reason. Just playing.
Thanks everyone for coming this weekend!
Best Regards, Martin Owens
Hey everyone,
Great meeting this Sunday
Event: https://inkscape.org/cals/event/64/
Attending: PBS, Martin, Elendil, Roland, René
Next meeting: June 14th (Sunday) at the later time
Link: https://inkscape.org/cals/event/1/
We talked about the CI building issues for windows briefly.
We talked about the bug accelerator program briefly.
Pbs asks about the capypdf static builds and the problem with debian
and macOS older versions. We conclude that the best was to tackle it is
to have the older versions detect C++23 and disable CapyPDF if not.
AppImage tooling issues have cropped up, pbs looked at mterry's fixes.
There's a OpenGL bug which causes AppImage to crash. There's a possible
solution to include their own version of mesa which is bloated but
might be able to fix it.
AppImages are mostly used for testing. Roland suggests using DistroBox.
Martin suggests using lazy AppImage building instead of building them
for every MR. Elendil says that while not useful for every commit it is
useful for merges into main for bisecting.
René, asks about using CI to get access to macOS for testing with
Roland, René will ask Roland in a few days to see how things are going.
Martin noticed a pattern with CI pipeine failures for project members.
PBS recommends using `git push -o ci.skip` to make sure your own GitLab
CI never runs and thus fails.
René reports he is burned out and will be taking a break for a few
months. Wants to make sure he doesn't leave Max G. with unfinished
packaging work, will message him in the issue.
Elendil has been looking at the new poppler release, more ifdefs. We
talked about the best ways of handling ifdefs in code, he's made compat
functions for things like pointers. We talked about how to talk with
poppler about a more long term solution with API breakge.
Roland talks about the bug accelerator program. Locked in on two
specific bugs right now, mterry is working on modifiers est at 10
hours. Seán is working on refactoring. The focus is on crashes and
making inkscape 1.5 more stable. Collecting the Gtk4 related issues
together for a sprint focused on that with things like menus, popovers
and similar.
We talk about the lot of Gtk3 issues that need to be closed. Should
they be tagged as gtk3 so they can be filtered out. we talk about 1.5
being a special case, while pbs makes the case for closing already
fixed issues.
Martin talked about their rendering engine work. Pbs asks about
rendering with Ints vs. Floats, martin says that having No color space
will use Ints and using RGB color space will use floats; to be
consistent with previous code. Martin says the code looks to be mostly
working how it intends to with some minor differences in rendering
output.
Pbs and Martin talk about code generation and testing, bounding boxes
for glyphs and making a new glyph drawing item for every single drawing
object. There's a bunch of work we can do to make text much much
faster.
Pbs finally asks about the shape builder task Seán is working on; asks
for any findings to be documented.
Best Regards, Martin Owens
