G
GTK's NSApplication delegate only implements application:openFiles:, so all URLs are dropped. On MacOS url schemas which are registered for the application and then clicked e.g. in a browser are passed as URL Event to applications. There is NSAppleEventManager where a URL handler needs to be registered for handleGetURLEvent. Its not much code for applications to implement themself, but GTK handles it on Linux and Windows, so would be nice to pass URLs also on macOS.
G
GTK's NSApplication delegate only implements application:openFiles:, so all URLs are dropped. On MacOS url schemas which are registered for the application and then clicked e.g. in a browser are passed as URL Event to applications. There is NSAppleEventManager where a URL handler needs to be registered for handleGetURLEvent. Its not much code for applications to implement themself, but GTK handles it on Linux and Windows, so would be nice to pass URLs also on macOS.
G
Replace application:openFiles: with application:openURLs:, which works for both files and URLs. URLs with a custom scheme were silently dropped before. Fixes: #8517 (closed) CC: @amolenaar
G
Steps to reproduce: Open gtk-demo Click on app menu Close app menu Click on app menu again Gdk-CRITICAL **: 22:32:07.193: gdk_surface_thaw_updates: assertion 'surface->update_freeze_count > 0' failed No Gdk-CRITICAL for the first popover open, after that, every time a popover gets opened. The same thing happens for tooltips, so it's likely related how surfaces freeze / thaw on macOS. General Information Name Value Operating System (null) GTK Version 4.25.0-31d14021 GDK Backend MacOS GSK…
G
@Jehan I tested the flatpak build and noticed that there is no Release Notes tab. I wonder if that is because this release does not have a list () on the metadata or it is a regression from the fix (1efebf2a) for Appstream/Flathub arbitrary linting
G
Yes, the Windows backend has the same issue. That's a bit surprising because https://docs.gtk.org/gdk3/struct.EventButton.html says that button is numbered from 1 to 5. I'll open an MR tomorrow!
G
Debian suggests that we can use an existing test file to do the test: https://salsa.debian.org/gnome-team/librsvg/-/blob/debian/latest/debian/patches/debian/tests-Replace-a-CC-BY-SA-2.0-test-image-with-a-simpler-on.patch
G
Marcus Lundblad (1aed368e) at 02 Nov 21:45 WIP: transitStopRow: Show timezone offsets ... and 1 more commit
G
Bruno (54e50113) at 02 Nov 21:40 build/linux: Update openexr module
G
AppKit reports the back and forward mouse buttons as 3 and 4, which were passed on to GDK as buttons 4 and 5. However, this does not match what Wayland or X11 do, which reserve buttons 4-7, and map extra buttons to 8 and up. This is the reason why the back button does nothing in widgets that listen for forward/back buttons, for example AdwNavigationView, which navigates back on button 8. Therefore, remap any additional buttons to 8 and above, so it matches X11/Wayland. This also matches how…
G
Affected version Mutter main (the code is the same in 51.1), and 50.1 with merge request !4806 (merged) backported. Bug summary For xdg_session_v1.remove_toplevel, xdg-session-management-v1 says: "Remove a specified surface from the session and render any related xdg_toplevel_session_v1 object inert. The compositor should remove any data related to the toplevel in the corresponding session from its internal storage." Mutter deviates from this in two ways: xdg_session_remove_toplevel() acts only…
G
gdk_paintable_new_from_bytes() was calling svg_to_texture() for symbolic SVGs with unsupported features, which goes through the snapshot/render-node path. This is not thread-safe, since the renderer has global mutable state (e.g. static caches in gsk_reload_font). When icon preloading calls this from thread pool workers, multiple threads end up using the renderer concurrently, causing crashes in _pango_cairo_font_private_get_scaled_font() due to cross-thread sharing of PangoFont objects via the…
G
Marge Bot (b23710f6) at 02 Nov 21:22 Merge branch 'text-icons-and-threads-omg' into 'main' ... and 1 more commit
G
Add-border filter, improved font-loading code, text tool improvement, bug fixes. The 3.2.8 looks great. Amazing work from the GIMP Team !!
G
The binaries I'm going to upload: AppImage: https://gitlab.gnome.org/GNOME/gimp/-/jobs/7304648/artifacts/browse/build/linux/appimage/_Output/ macOS: https://gitlab.gnome.org/GNOME/gimp/-/jobs/7304652/artifacts/browse/build/macos/dmg/_Output/ The Windows installer which has to be signed by @jernejs: https://gitlab.gnome.org/GNOME/gimp/-/jobs/7304650/artifacts/browse/build/windows/installer/_Output/ The MS Store and Snap should be automatically updated. This is the latest test Flatpak: flatpak…
G
Jens Georg (45912153) at 02 Nov 20:58
G
tile_window() sets the tile mode on window->config before set_maximize_state() saves the rect, so save_rect() sees a non-floating window and returns early. Keyboard tiling passes no explicit rect, so saved_rect is never written and untiling uses an empty rect (videos in #4918). Wayland windows are affected too: they keep their size but stay at the tiled position. This saves the rect before the tile mode is changed. An explicit rect (mouse drag) is still used as before. The toggle-tiled…
G
Good news - it seems like the cases openQA was running into got fixed in GNOME 50 Final, I think. The test has passed on aarch64 every day since then. I also randomly checked a couple of tests and didn't see any where it failed to switch on the first combo press and relied on the retry mechanism. So, I think this is fixed. Thanks!
G
I was thinking something like: pin two threads (on select kernels, give them different priorities) to a single CPU core, have them yield to each other and set an "I've been there" flag (increment an atomic). We then check both that this doesn't take overly long (so no-one slept), and that between calling yield and it returning, the other thread got a chance to make some progress (say, at least in 90% of iterations).
G
This was only a temporary workaround for GNOME 50 (51) and should not be in later versions. Please merge after gnome-51 is branched. Thanks!
G
Marge Bot (e5582f25) at 29 Oct 11:01 data: Remove old Passwords and Keys desktop name
G
Introduce AGO joining and capability helpers without requiring extended libnm APIs for ordinary PBC connections. Assisted-By: Codex noreply@openai.com Needs updated NetworkManager, pending merge request here: https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/2585
G
Thanks @ignapk for making this fix. Would this be backported to the gnome-51 branch (51.2)?
G
Marge Bot (6e156b2e) at 29 Oct 10:57 fileUtils: Don't require deleted dir to exist ... and 3 more commits
G
Change recursivelyMoveDir() to not leave directories around. Change recursivelyDeleteDir() to handle NOT_FOUND errors if the directory itself should be deleted. Make use of Gio.FileEnumerator iterators.
G
Jehan (1a6f49a2) at 29 Oct 10:53 meson, INSTALL: do not include comparator sign in the version.
G
Hi, sorry, I don't know more than what's documented, and this ticket has been closed in favor of gweather-locations#70 :)
G
Bumped to allow 100us/call as a cutoff, as the macos-arm64 and FreeBSD CI runners average 27us/call and 58us/call respectively.
G
Maybe a post in https://discourse.gnome.org/tag/evolution/ or on the mailing lists could bring some more input
G
Philip Withnall (409e3f2a) at 29 Oct 10:47 tests: Add a performance regression test for g_thread_yield()