This is a possible fix for https://gitlab.gnome.org/GNOME/gtk/-/issues/2657
Use a NULL return from g_file_query_info_finish() to detect cancellation
of the query, and avoid derferencing a stale pointer.
GDK has a lock to mark critical sections inside the backends.
Additionally, code that would re-enter into the GTK main loop was
supposed to hold the lock.
Back in the Good Old Days™ this was guaranteed to kind of work only on
the X11 backend, and would cause a neat explosion on any other GDK
backend.
During GTK+ 3.x we deprecated the API to enter and leave the critical
sections, and now we can remove all the internal uses of the lock, since
external API that uses GTK+ 4.x won't be able to hold the GDK lock.
https://bugzilla.gnome.org/show_bug.cgi?id=793124
The main GDK thread lock is not portable and deprecated.
The only reason why gdk_threads_add_timeout() and
gdk_threads_add_timeout_full() exist is to allow invoking a callback
with the GDK lock held, in case 3rd party libraries still use the
deprecated gdk_threads_enter()/gdk_threads_leave() API.
Since we're removing the GDK lock, and we're releasing a new major API,
such code cannot exist any more; this means we can use the GLib API for
installing timeout callbacks.
https://bugzilla.gnome.org/show_bug.cgi?id=793124
This patch makes that work using 1 of 2 options:
1. Add all missing enums to the switch statement
or
2. Cast the switch argument to a uint to avoid having to do that (mostly
for GdkEventType).
I even found a bug while doing that: clearing a GtkImage with a surface
did not notify thae surface property.
The reason for enabling this flag even though it is tedious at times is
that it is very useful when adding values to an enum, because it makes
GTK immediately warn about all the switch statements where this enum is
relevant.
And I expect changes to enums to be frequent during the GTK4 development
cycle.
When the search engine provides hits with GFileInfo, use that
to add the hits to the model directly, without going through
another round of async get_info calls.
To do this, we add a batched variant of the
_gtk_file_system_model_update_file call that takes lists of
GFiles and GFileInfos. Again, we can avoid repeated resorting
that happens when the files are updated individually.
Add a batched version of gtk_file_system_model_add_and_query_file
that takes a list of files and avoids resorting the model for each
individual insertion. The querying is still done one-file-at-a-time,
so more optimization is certainly possible.
If a file system event arrives while GtkFileChooserWidget is asking the
user to edit the name of a newly created folder, the file system model
will drop the row with the editable cell, and the user will have to
start from scratch.
This makes creating new directories impossible inside a directory with a
file currently being downloaded, for instance, and it's really unhelpful
to the user because the editable row simply disappears.
We already have a mechanism in place to freeze the file system model, so
we can reuse it between the add_editable() and the remove_editable()
calls.
https://bugzilla.gnome.org/show_bug.cgi?id=729927
The main problem is that we were emitting the row-deleted signal for the model in the middle
of the process that actually deletes the row from the model (remove the row from the array,
update the model->file_lookup hash table, etc.). In the model's caller, one of the row-deleted
callbacks was requesting an iter, which caused the model to revalidate itself - but it did
this while it was in an inconsistent state. This led to an assertion failure later when the
model resorted itself.
The fix in remove_file() is like this:
* The filteredness/visibility of the deleted node is not updated. The
node will simply be gone; we don't need to update those values at
all.
* We invalidate just the node that is being deleted.
* The model->file_lookup hash table is not completely nuked; instead,
we carefully adjust its indices.
* The row-deleted signal is only emitted at the very end, when
deletion is complete and the model is consistent.
Many thanks to William Hua for doing the detective work on this bug!
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
They were in the semi-public API of GtkFileSystemModel, but never actually used outside of it.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This is a function internal to the file system model; let's not pollute the gtk_tree_path namespace.
Also, make the 'i' variable into 'r' as it refers to a row index, not a file-array index (for
consistency with the docs and the rest of the code).
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This way we remove paired function calls (compute/set pairs), and also make
it possible to avoid computing a filter twice, as setting the visibility
depends on filteredness.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
It bothers me that we call gtk_file_filter_filter(), then negate the result,
and the return *that* from node_should_be_filtered(). So, rename 'filtered'
throughout GtkFileSystemModel to 'filtered_out' to mean things that didn't
pass the filter.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This was a copy-paste leftover from node_set_visible(). Filters are not
concerned with model freezes, so node_set_filtered() does not
need to handle freezes, either.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Async callbacks are delivered in idles, so we need to make sure
we get the gdk lock before calling any gdk/gtk stuff. This was
missing in a few places.
When a file was inserted during the period that the editable row was
active, the node IDs would not get updated correctly.
Signed-off-by: Federico Mena Quintero <federico@novell.com>
The old version wasn't introspectable as it didn't have a length
return parameter. Also, delete gtk_tree_path_get_indices_with_depth,
since it's no longer needed.
The file removal code was not properly clearing the file=>array index
cache, so later lookups into that cache would return invalid array
indexes.
The easiest way to reproduce it is to create a directory with two files
and deleting both of them.
Reported-by: Javier Jardón <jjardon@gnome.org>
We should not unref the model here, it might not even exist anymore.
Instead check if it exists and only use it if it does.
The unref was leftover from a previous fix in
ba9f53397f.
Spotted by Matthias Clasen in
https://bugzilla.gnome.org/show_bug.cgi?id=614099
This commit was created using a script that searched for all docstrings
containing a parameter and the string 'or %NULL'.
Gdk backends and demos excluded as they are not part of a public API
https://bugzilla.gnome.org/show_bug.cgi?id=610474
This allows disposing of the filesystemmodel while the file enumeration
is still happening.
As the filechooser does not disconnect its signals because it assumes it
is the only owner of the model, this also prevents a SEGV when emitting
the "load-finished" signal in that case.
