Basically disabling commit 4f9b7373e6.
After some new patches for GTK3 I wrote, and removing the settings on
NSViewUsesAutomaticLayerBackingStores, Lukas reported that it works like
never before, faster than 2.10 even. Note that this could only be tested
on Monterey, nobody tested on Big Sur with this specific combination
yet.
In any case, we decided to remove this settings and add the new GTK3
patches.
See: https://gitlab.gnome.org/Infrastructure/gimp-macos-build/-/merge_requests/86#note_1384727
It's a bit ugly, but it's not like this is run many times (only once
when loading the icon theme, or changing it).
Fixes this error appearing in various unit tests' output:
> gimp_icons_sanity_check: Icon theme path has no 'hicolor' subdirectory: /builds/GNOME/gimp/_install/share/gimp/2.99/icons
(even though it was not a test-failure error, it's still better to limit
output for debugging)
When running tests, the data are not meant to be necessarily installed.
Therefore icons won't be found when calling gimp_widgets_init().
Add some special-casing to find them relatively to the install
directory.
Reviewer comment (Jehan): we have used this patch successfully on our
installers since start of 2021 (see commit b4d665d of our gtk-osx fork)
and it really improved the situation. I only fixed minor coding style
stuff in the patch.
Looking at what it does, I guess it is not ideal long-term if related to
10-bit display (as I understand from the comment), which a graphics app
would want to support properly. But for now, this is better than
extra-slow display until we get macOS developers able to look at this
more in depth in the future (I don't think that our dependencies are
really ready yet for 10-bit display support anyway, though I may be
wrong).
Some other forums seem to say it comes from macOS invalidating now more
than it should (i.e. the whole area instead of only the changed area)
and this NSViewUsesAutomaticLayerBackingStores flag would disable this
behavior. It might be one of these reasons, the other or both. This is
anyway a good first start for future contributors.
Simply it should free only GimpProgressDialog as these would be
dedicated dialog (with no meaning once progression is done), and leave
alone other GimpProgress widgets. In any case, it should not output
CRITICAL errors on these.
Fixing the following CRITICAL:
> GIMP-CRITICAL: gui_free_progress: assertion 'GIMP_IS_PROGRESS_DIALOG (progress)' failed
When dropping an image on the toolbox.
In some cases, error dialogs may end up below other windows and stay
unnoticed because of how dialogs are raised and hidden on various
platforms. This is not very constructive. It's much better to make sure
one sees an error when it happens (in some cases, it may mean possible
data loss, so it should be at least acknowledged; also seeing it later
may mean you can't know anymore which action triggered this error,
making the whole process kind of meaningless and hard to debug).
So anyway, let's make error messages prominent by having them always
above.
- Set the software as `initialized` later, and in particular after all
recovered images (from crash) then all command line images were
opened. The reason is that the DBus calls have necessarily been made
after GIMP was started (typically could be images double-clicked
through GUI). We don't want them to appear before the images given in
command line (or worse, some before and some after).
- Process DBus service's data queue as a FIFO. The image requested first
will be loaded first.
- When a DBus call happens while GIMP is not initialized or restored,
switch to a timeout handler. The problem with idle handlers is that
they would be attempted too often (probably even more during startup
when no user event happens). This is good for actions we want to
happen reasonably quickly (like would be normally DBus calls), but not
when we are unsure of program availability schedule (i.e. at startup).
Here not only the handler would run a lot uselessly but it would
likely even slow the startup down by doing so. So while GIMP is not
initialized, switch to half-a-second timeout handler, then only switch
back to idle handler when we are properly initialized and GIMP is
ready to answer calls in a timely manner.
… DBus calls.
In particular, Aryeom would start GIMP and directly double click some
image to be loaded in GIMP in the very short while when splash is
visible. Previous code would wait for the `restored` flag to be TRUE.
This was nearly it as we can actually start loading images as soon as
the 'restore' signal has passed. Yet the flag is set in the main
handler, but we actually also need the <Image> UI manager to exist,
which is created in gui_restore_after_callback() (so also a 'restore'
handler, yet after the main signal handler, i.e. after `restored` is set
to TRUE). Without this, gui_display_create() would fail with a CRITICAL,
hence file_open_with_proc_and_display() as well.
I could have tried to set the `restored` flag later, maybe with some
clever signal handling trick (and handle both the GUI and non-GUI cases,
i.e. I cannot set the flag inside gui_restore_after_callback() as it
would break the non-GUI cases). Instead I go for a simpler logics with a
new `initialized` flag which is only meant to be set once, once
everything has been loaded, i.e. once you can consider GIMP to be fully
running hence ready to process any common runtime command.
This is not a fix, only an extra-ugly workaround so that at the very
least we don't end up with a splash screen taking the whole display on
Wayland.
Basically by setting 1/3 as the max splash size, a Wayland desktop with
no scale ratio will have a splash taking a third of the screen while it
would take 2/3 of the screen with a scale ratio of ×2 (of course, it
will still be very broken with a scale ratio of ×3 but are there
displays needing such high scaling?). The real fix will be when GTK/GDK
fix their API so that it returns what the docs says it should (i.e. a
size in "application pixels" not "device pixels"), as it does on X11,
Windows, macOS… Then we won't create random max size and we will be able
to properly control our splash size.
Note that this neither fixes nor works around the position issue on
Wayland (in my case, the splash was just always on top-left of the
display).
These small glitches have bothered me for a while now, so I finally
fixed these before the dev release!
Basically there were 2 fixes:
1. use the ink extents to compute any drawn area as this is what will be
actually drawn.
2. Not only expose the drawn area of the new text, but also the one of
the previous text in order to be sure all text pixels are correctly
reset (in case the new text is smaller than previous one). I.e. we
must expose the smallest rectangle containing both previous and new
area of text.
When loading a theme on Windows we always get an error like:
themes_theme_change_notify: error parsing [path including drive letter to:]\theme.css: theme.css:8:99Failed to import: Operation not supported
The location points to the end of the filename of the first @ import url string.
This is caused by the string not being an url.
Based on suggestions from Jehan and lillolollo we replace g_file_get_path
with g_file_get_uri since an url is what is expected here.
Since that function already escapes the string we also remove
g_str_escape here.
Orientation is now handled by core code, just next to profile conversion
handling.
One of the first consequence is that we don't need to have a non-GUI
version gimp_image_metadata_load_finish_batch() in libgimp, next to a
GUI version of the gimp_image_metadata_load_finish() function in
libgimpui. This makes for simpler API.
Also a plug-in which wishes to get access to the rotation dialog
provided by GIMP without loading ligimpui/GTK+ (for whatever reason)
will still have the feature.
The main advantage is that the "Don't ask me again" feature is now
handled by a settings in `Preferences > Image Import & Export` as the
"Metadata rotation policy". Until now it was saved as a global parasite,
which made it virtually non-editable once you checked it once (no easy
way to edit parasites except by scripts). So say you refused the
rotation once while checking "Don't ask again", and GIMP will forever
discard the rotation metadata without giving you a sane way to change
your mind. Of course, I could have passed the settings to plug-ins
through the PDB, but I find it a lot better to simply handle such
settings core-side.
The dialog code is basically the same as an app/dialogs/ as it was in
libgimp, with the minor improvement that it now takes the scale ratio
into account (basically the maximum thumbnail size will be bigger on
higher density displays).
Only downside of the move to the core is that this rotation dialog is
raised only when you open an image from the core, not as a PDB call. So
a plug-in which makes say a "file-jpeg-load" PDB call, even in
INTERACTIVE run mode, won't have rotation processed. Note that this was
already the same for embedded color profile conversion. This can be
wanted or not. Anyway some additional libgimp calls might be of interest
to explicitly call the core dialogs.
While this issue was unseen so far on common desktop machines, the CI
build encountered it, hence failing 6 of the unit tests.
A connection to the bus could not be established hence the dbus_manager
was never allocated, and finally it would crash at exit if we tried to
unref it unconditionnally. Use g_clear_object() instead.
Also add some stderr output for easier debugging, for when one of the 2
possible error cases might happen (as documented by g_bus_own_name()).
Tested in a VM. Minimized window is properly deiconified and showed to
the front. Though a window in the back (not minimized) is not moved to
the front.
GimpDisplay contains only the ID logic and the "gimp" and "config"
pointers, and lives in the core.
GimpDisplayImpl is a subclass and contains all the actual display
stuff. The subclass is only an implementation detail and doesn't
appear in any API.
Remove all hacks which pass displays as gpointer, GObject or
GimpObject through the core, or even lookup its type by name,
just use GimpDisplay.
Turn all ID param specs into object param specs (e.g. GimpParamImageID
becomes GimpParamImage) and convert between IDs and objects in
gimpgpparams.c directly above the the wire protocol, so all of app/,
libgimp/ and plug-ins/ can deal directly with objects down to the
lowest level and not care about IDs.
Use the actual object param specs for procedure arguments and return
values again instead of a plain g_param_spec_object() and bring back
the none_ok parameter.
This implies changing the PDB type checking functions to work on pure
integers instead of IDs (one can't check whether object creation is
possible if performing that check requires the object to already
exist).
For example gimp_foo_is_valid() becomes gimp_foo_id_is_valid() and is
not involved in automatic object creation magic at the protocol
level. Added wrappers which still say gimp_foo_is_valid() and take the
respective objects.
Adapted all code, and it all becomes nicer and less convoluted, even
the generated PDB wrappers in app/ and libgimp/.
GimpConfigWriter contains several constructors with the convention
`gimp_config_writer_new_* ()`. This will lead to problems however with
languages like Vala, where it cannot disambiguate the following:
```
// calls config_writer_new_string()
Gimp.ConfigWriter w = new ConfigWriter.string("xxx");
// calls config_writer_string()
w.string("xxx")
```
Using `from_` in constructors is general practice in GObject-bsed
libraries because of this.
This also fixes an error when trying to use vapigen on the GIMP .GIR
file.
which means that it's now included normally via gimpbase.h
and not any longer via gimpbasetypes.h which we only did out
of lazyness. A *lot* of files in libgimp* and app/ now need to
- libgimpbase: change GPParam to transfer all information about the
GValues we use, in the same way done for GPParamDef. GPParam is now
different from GimpParam from libgimp, pointers can't be casted any
longer. The protocol is now completely GimpPDBArgType-free. Remove
gp_params_destroy() from the public API.
- libgimp: add API to convert between an array of GPParams and
GimpValueArray, the latter is now the new official API for dealing
with procedure arguments and return values, GimpParam is cruft (the
wire now talks with GimpPlugIn more directly than with the members
of GimpPlugInInfo, which need additional compat conversions).
- libgimp, app: rename gimpgpparamspecs.[ch] to simply
gimpgpparams.[ch] which is also more accurate because they now
contain GValue functions too. The code that used to live in
app/plug-in/plug-in-params.h is now completely in libgimp.
- app: contains no protocol compat code any longer, the only place
that uses GimpPDBArgType is the PDB query procedure implementation,
which also needs to change.
- app: change some forgotten int32 run-modes to enums.
There are no replacements. Just we must make sure that all GTK+/GDK
calls are run from the main thread, which is already what we were doing.
Actually I don't even think these were doing anything as we were not
calling gdk_threads_init() so the default lock functions were not set
anyway. These were just bogus calls.
Step one: get rid of all those deprecation warnings that make
it hard to see any other warnings:
- add a lot of dummy API to GimpAction, GimpActionGroup, GimpUIManager
etc. which simply forwards to the deprecated GTK functions, they
will all go away again later
- rename GimpAction to GimpActionImpl
- add interface GimpAction that is implemented by all action classes,
creates a common interface and allows to remove some duplicated
logic from GimpToggleAction and GimpRadioAction, and at the same
time adds more features
Include the system-wide gimp.css file, in addition to the user-
specific gimp.css file, in the generated theme.css file, instead of
copying the former into the latter when creating the user's
gimpdir. This allows us to modify the system-wide gimp.css file,
and having the changes take effect in existing installations.
The GUI implementation of gimp_wait() relies on the ability to run
plug-ins (namely, the busy-dialog plug-in) without entering the
main loop. This prohibits the said plug-ins from making any PDB
calls, which would result in a deadlock. However, we're currently
calling gimp_gimprc_query() to fetch the prefer-dark-theme option
during gimp_ui_init() (or any time the theme.css file changes).
Instead, communicate this preference through the theme.css file
itself, by writing a /* prefer-dark-theme */ comment to the file
when the option is set. Yes, it's a bit of a hack :P
When an error occurs, we want to prevent overwriting any previous
version of the file by incomplete contents. So run
g_output_stream_close() with a cancelled GCancellable to do so.
See also discussion in #2565.
I had this funny behavior when I was quitting GIMP with the active tool
using modifiers (for instance bucket fill). Each time I'd quit with
ctrl-q (and if the image is not dirty), the options would use the value
from the modifier state and be saved as-is. Hence at next restart, the
default value was always different!
In the GUI implementation of gimp_wait(), explicitly finish the
input-pipe async operation after the busy-dialog plug-in
terminates, to avoid the async callback function from being
repeatedly called, stalling the main thread. Previously, this code
relied on gimp-parallel implicitly aborting the async operation,
but this is no longer the case since commit
4969d75785.
Anders Jonsson