These were originally created to better support alpha/transparency
in slideshows, but vcl can do that these days.
Mostly, removing these makes the path to better backends (e.g. GDI+)
in vcl considerably easier.
If regressions are detected, it would be better to improve vcl
to perform better, or render better, since that will both fix
the regression and improve other output from LibreOffice.
Also remove various bits of vcl backend that were only being
used by the canvas stuff.
Change-Id: Ic5276d57e696f3dcbeaed2cc149b8fc09c5816f6
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/181843
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
This has not been necessary for some years now. In the unlikely
event that we run on a 256-color display, the OS and the graphics
driver will do the dithering for us.
Change-Id: I23c7a3ded6ea63fcd0a5239bbf2a2cc91bba306f
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/181143
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
regression from
commit e0d4d178ca
Author: Noel Grandin <noel.grandin@collabora.co.uk>
Date: Mon Jan 13 15:03:05 2025 +0200
Change alpha behavour of OutputDevice::SetFillColor
where DrawColorWallpaper is not correctly saving/restoring the
fill color.
Previously this worked because GetFillColor would return a
transparent color if no fill color was set, and a transparent
fill color would be interpreted as "no color" when drawing.
Change-Id: I63e1ba2df0a5e2088d018068fe0a0e45453b7c77
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/180928
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
One of the paths cannot be hit, so remove it.
Add asserts in other places to catch conditions that
should not happen.
Change-Id: I401a0ffea2587d8bbcb313edef06e766d65c3eee
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/180533
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
Tested-by: Jenkins
regression from commit 21734247d5
"drop internal support for 1-bit images"
This only works on Linux because the HasGreyPaletteAny() check
is returning false on Linux, because on linux the swap in/out is more aggressive, and after it is has gone through the swap in/out process,
the image palette is no longer one that matches anything that HasGreyPaletteAny() checks for (it contains 256 entries, but only 2 of them are used)
I'm not exactly sure why the greyscale bitmap generation code
is generating bad PDF data, but the RGB code works great, so
lets just use that rather.
Change-Id: Ibb9e837540d5ed567c706e21e7ff93fe92118580
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/177329
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
so I messed up and uploaded skia m117 twice. But now there are two
copies of skia m117 in various CDN caches, with different sha256
output (no idea how that happened).
So upload a THIRD COPY with a new name, and then I can fix the
sha256 value in download.lst in a way that doesn't break the build for
someone.
Change-Id: I297ff4d9fa3a0edd902c85f2d674950945249a5e
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/174590
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
while testing the document attached to this bug I get a repeatable
skia crash (only skia is in native mode, not software rendering).
This reveals that there a couple of data structures that need to
be cleaned up while holding the SolarMutex.
Change-Id: I3ff460a4febf1ffcb27ad904224e1594ead4aa3d
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/170612
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
My debug build is slow enough already, no need to make it worse.
People who want it, can turn it on explicitly.
Change-Id: I8677534d8f0142699baa6b95a249ae5f70c5cc3c
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/170269
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
after
commit 4c86718e78
Author: Christian Lohmaier <lohmaier+LibreOffice@googlemail.com>
Date: Thu Apr 18 12:45:01 2024 +0200
initial support for running autogen.sh inside wsl from git-bash
otherwise we end up with mixed cygwin/windows paths in our .d
files, which the make executable we use will not like.
Change-Id: Ia1325793f47657a23774c216df319ae6afd6d638
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/167269
Tested-by: Jenkins
Reviewed-by: Christian Lohmaier <lohmaier+LibreOffice@googlemail.com>
after
commit defc7a9206
Author: Noel Grandin <noel.grandin@collabora.co.uk>
Date: Fri May 3 09:20:04 2024 +0200
replace createFromAscii with OUString literals in vcl::pdf::BuildinFont
my local version of Visual Studio seems to have better support
for constinit than the jenkins one.
Change-Id: If6aec53e8280c5afbd27bf53b7ac0cd1a5e742b0
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/167185
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
std::multiset seems to be pretty slow to construct and
destruct, even if we ~never put anything into these listener
lists, so rather use o3tl::sorted_vector, which has more
predictable performance.
Shaves 10% off startup time for me.
Change-Id: I13afcbe3cb63522e4efe61c64f773e1ffbec9683
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/167101
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
Comparing with zero is simple - the implementation of basegfx::fTools::moreOrEqual
calls rtl_math_approxEqual eventually, which special-zases zero.
Change-Id: I62f10f63f103d91a201dfeb20e5b3f9010f377c1
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/164526
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
after
commit 191e3f8d89
Author: Noel Grandin <noelgrandin@gmail.com>
Date: Mon Feb 19 15:48:38 2024 +0200
merge textconv_dict into i18npool
The first build would work, but subsequent builds would fail.
I'm not sure why this patch fixes it, but it does. Somehow the sed
command was affecting other generated files?
Change-Id: I1550709b6555c838c0371e31b5eb2c90f5065e10
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/163648
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
revert 69910b540a and
add a couple of pre-emptive reserve calls, to prevent repeated
resizing.
I'm not sure why the above commit causes trouble on Windows, but
not Linux, something in the std::vector::reserve implementation
no doubt.
Change-Id: I858303a0a1e12d204fd3bbccc6c6c7ce57564e5b
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/162746
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
Tabbed pane titles are currently incredibly cramped,
increase the space a little to make them match better
with current OS layouts (all of the different OSes
we support use quite a lot of space around the
titles)
Change-Id: If9c7b4df81f8d206556e7c75097650ae21508384
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/153733
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
I want to know the symbol address we actually capture, since
we seem to getting duplicates in the output of SymFromAddr, I
want to know if the fault is in SymFromAddr, or the backtrace
capture call.
Change-Id: Iece9c204abb780ae6b67a11e6ba77db816351e8b
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/151254
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
repeated calls to resize() means we bypass the normal doubling behaviour
of the vector, so rather let it do it's own thing.
Change-Id: I7b4273d20b6cdcac6aeddd61395fbe64de004c9f
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/142176
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
Noel Grandin