tests: Mark gltexture as expected to fail on big-endian machines

Forwarded: not-needed, workaround

Gbp-Pq: Topic debian
Gbp-Pq: Name tests-Mark-gltexture-as-expected-to-fail-on-big-endian-ma.patch

gstreamer-player: Use GtkName Prefix to avoid issues if actual library is included

Origin: https://gitlab.gnome.org/3v1n0/gtk/-/commits/gst-player-subproject
Forwarded: not-needed

Gbp-Pq: Topic debian
Gbp-Pq: Name gstreamer-player-Use-GtkName-Prefix-to-avoid-issues-if-ac.patch

media: Use gstplayer from meson subproject

Origin: https://gitlab.gnome.org/3v1n0/gtk/-/commits/gst-player-subproject
Forwarded: not-needed

Gbp-Pq: Topic debian
Gbp-Pq: Name media-Use-gstplayer-from-meson-subproject.patch

gstreamer-player-1.0: Do not install it if it's a subproject

Origin: https://gitlab.gnome.org/3v1n0/gtk/-/commits/gst-player-subproject
Forwarded: not-needed

Gbp-Pq: Topic debian
Gbp-Pq: Name gstreamer-player-1.0-Do-not-install-it-if-it-s-a-subproje.patch

subprojects: Add gst-player-1.0 as subproject

Origin: https://gitlab.gnome.org/3v1n0/gtk/-/commits/gst-player-subproject
Forwarded: not-needed

Gbp-Pq: Topic debian
Gbp-Pq: Name subprojects-Add-gst-player-1.0-as-subporject.patch

Disable inscription-markup.ui reftest

It's rendered 1px to the left of where it should be, which is not a
serious problem for practical use of GTK, but is a test failure.

Bug: https://gitlab.gnome.org/GNOME/gtk/-/issues/5099
Forwarded: not-needed, workaround

Gbp-Pq: Topic debian
Gbp-Pq: Name Disable-inscription-markup.ui-reftest.patch

reftests: Allow minor differences to be tolerated

Based on an earlier patch by Michael Biebl, as used in Debian's GTK 3
packaging, with additional inspiration from librsvg's reftests.

Each .ui or .node reftest can have an accompanying .keyfile file
like this:

    [reftest]
    tolerated-diff-level=20
    tolerated-diff-pixels=1000

If the image differs, but the number of pixels that differ is no more
than tolerated-diff-pixels and the differences are no more than
tolerated-diff-level, then we treat it as a success with warnings, save
the .diff.png for analysis, and use g_test_incomplete() to record the
test-case as "TODO".

Signed-off-by: Simon McVittie <smcv@debian.org>
Forwarded: https://gitlab.gnome.org/GNOME/gtk/-/merge_requests/3195
Applied-upstream: no, upstream want reftests to be a strict pass/fail with identical results required

Gbp-Pq: Topic debian
Gbp-Pq: Name reftests-Allow-minor-differences-to-be-tolerated.patch

reftest_compare_surfaces: Report how much the images differ

In unattended/non-interactive/autobuilder environments where the images
are not trivially accessible, this provides a way to distinguish between
totally different rendering and more subtle issues.

Signed-off-by: Simon McVittie <smcv@debian.org>
Forwarded: https://gitlab.gnome.org/GNOME/gtk/-/merge_requests/3195
Applied-upstream: no, upstream want reftests to be a strict pass/fail with identical results required

Gbp-Pq: Topic debian
Gbp-Pq: Name reftest_compare_surfaces-Report-how-much-the-images-diffe.patch

tests: mark headless tests as flaky

See baaa7482489b0f7446ef1e0130e527a169c43b0c

Gbp-Pq: Name tests-mark-headless-tests-as-flaky.patch

Revert "build: Drop the install-tests option"

Debian and Ubuntu uses these installed tests as part of their
autopkgtest system

This reverts commit 3121f88265ac61733e257f5335473d6f139f158c.

Gbp-Pq: Name Revert-build-Drop-the-install-tests-option.patch

Revert "tests: Stop copying the tool tests"

This reverts commit bb3a6214cbeaa3211be007192e536bd473323836.

This is actually used by the installed tests
which we also had to revert dropping (in a separate patch)

Gbp-Pq: Name Revert-tests-Stop-copying-the-tool-tests.patch

gsk: Always use RGBA on big-endian architectures

The memorytexture test started failing on big-endian CPUs while using
"desktop" GL (not GLES) in commit b0e26873 "gsk: Use has_bgra in more
places". The subsequent commit f5812808 "gsk: Restore bigendian support"
was meant to fix this, but it seems it did not.

Always taking the RGBA code-path on big-endian might not be optimal,
but it seems to result in the test passing again.

Signed-off-by: Simon McVittie <smcv@debian.org>
Bug: https://gitlab.gnome.org/GNOME/gtk/-/issues/6260
Bug-Debian: https://bugs.debian.org/1057782
Forwarded: https://gitlab.gnome.org/GNOME/gtk/-/merge_requests/6653

Gbp-Pq: Name gsk-Always-use-RGBA-on-big-endian-architectures.patch

gtk4 (4.12.5+ds-2) unstable; urgency=medium

  * Allow stderr for python3-gi autopkgtests to not fail
    tests for zink warnings emitted by mesa

[dgit import unpatched gtk4 4.12.5+ds-2]

Import gtk4_4.12.5+ds-2.debian.tar.xz

[dgit import tarball gtk4 4.12.5+ds-2 gtk4_4.12.5+ds-2.debian.tar.xz]

Import gtk4_4.12.5+ds.orig.tar.xz

[dgit import orig gtk4_4.12.5+ds.orig.tar.xz]