| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
| |
If we don't own the buffer, then we don't have a guarantee that the buffer
is one byte bigger than any count/length we've received. Mke sure to only
dereference that byte when we do.
Related #309
|
| |
|
| |
|
| |
|
|
|
|
| |
This was changed to be buffer coordinates and should be replicated here.
|
|
|
|
|
| |
This had lots of changes in master, so we should pin to the release that
matches this version of GtkSourceView.
|
|\
| |
| |
| |
| | |
build: Also try finding libxml2 with CMake
See merge request GNOME/gtksourceview!267
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
For Windows builds of libxml2, one may be using its supplied NMake Makefiles
to build libxml2, which do not have complete pkg-config files, or use CMake,
which optionally crafts its pkg-config files that may need some tinkering for
proper use.
Instead, save from such issues to try finding again libxml2 with CMake, which
actually manually search for libxml2's headers and libraries like what we used
to do, for us, if a proper pkg-config file cannot be found by Meson, before we
resort to the fallback.
If at one point we require Meson 0.60.0 (or so) or later, we can just put the
package strings that we want to search for in a form like
['libxml-2.0', 'LibXml2'], but let's just do this for the time being.
The "LibXml2" format follows the CMake docs at [1].
[1]: https://cmake.org/cmake/help/latest/module/FindLibXml2.html
|
|
|
|
| |
See NEWS for release information.
|
| |
|
|
|
|
|
|
|
|
| |
That callback is not necessarily needed in many cases and will cause
signal emission (for better compatibility with older GtkSourceView).
So check for NULL and skip that in the hot path for cases where we know
it won't be used such as some high-priority subclasses.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
There were two issues:
- The first regex was matching =end as well
- The wrong capture variable was being referenced in the =end regex
These led to =end being completely ignored, causing syntax highlighting
to cease for anything after the initial =begin (including =cut).
This also modifies the file.pl test cases to include a section of POD
that'll break highlighting without this commit applied. An example of
__END__ was added as well. The POD in the file additionally now renders
properly with 'perldoc'.
|
|
|
|
|
|
|
|
|
| |
This makes things look somewhat similar to Solarized patch viewing in
other editors.
The change-line is kept yellow instead of orange since that isn't really
used for display anywhere notable, yet can be used by IDE tooling for
the gutters and orange is too close to red there.
|
|
|
|
|
| |
- Update to Fedora 36
- Add some additional packages to reduce what needs to be built
|
|
|
|
| |
And avoid if there is no surface yet or not visible.
|
|
|
|
|
|
|
|
| |
And update the informative immediately afterwards as the visible rect
will have changed.
This doesn't use jump_to_iter because we don't care about any particular
alignment and this seems to get the job done more correctly.
|
|
|
|
| |
Otherwise we'll only place things correctly on the first visible page.
|
|
|
|
| |
These appear to be unnecessary.
|
|
|
|
|
| |
This fixes GtkSourceStyleSchemePreview so that it doesn't create a new
informative for every widget. GtkSourceMap as well.
|
| |
|
|
|
|
|
| |
Now that we (more correctly) have vim registers shared between GtkSourceVim
stacks, we need to reset them before each test case.
|
|
|
|
|
|
|
|
| |
So that they can be shared between instances of GtkSourceView with a vim
controller attached.
This moves the state global rather than the register object global which
caused issues with leave/enter semantics.
|
|
|
|
| |
We want to use this internally with tests.
|
|
|
|
|
|
|
|
| |
The two strv should be compared in such a way that the complete set is in
both strv but in any order and no more. Just a general set theory
comparison of the two strv.
Based on suggestion from Sébastien Wilmet on discourse.gnome.org.
|
|
|
|
| |
This just results in doing less work while the widget is already hidden.
|
|
|
|
|
| |
This turns out to happen more than anticipated, and could cause us to get
into a situation where we loose allocations on the source view.
|
|
|
|
| |
But only when the offset will not change (added == removed).
|
|
|
|
|
|
| |
Sometimes GtkTextView will cause us to get notifications when things didn't
really change. This protects against that with a cached value so that we
can more readiliy avoid extraneous size allocations.
|
| |
|
|
|
|
| |
This will dismiss a pending hover if the user moves the cursor first.
|
| |
|
|
|
|
| |
Related #265
|
| |
|
| |
|
| |
|
|
|
|
| |
This was missing from the implementation by accident it appears.
|
| |
|
|
|
|
|
|
| |
This reverts commit 67d6e60d7382cc3236825749b272e1dd7489bdf9.
This shouldn't have been backported.
|
| |
|
|
|
|
|
|
| |
This ensures we go through a resize-cycle when the right-margin-position
is updated in the linked GtkSourceView. Otherwise our size will not update
with the linked view.
|
|
|
|
|
| |
Make it more flexible so apps can e.g. add .card and get libadwaita styles
on it.
|
| |
|
|
|
|
|
|
| |
It can be helpful to have an extra line before/below when rendering
gutter contents so that you can fill in information you need based
on a single neighboring line.
|
|
|
|
| |
Fixes #263
|
|
|
|
|
|
|
|
| |
Rather than hardcoding specialized values in fallback schemes, we should
generate better values automatically as some schemes in the whild will be
missing these values anyway.
Fixes #260
|
|
|
|
|
|
| |
This code existed before the simplified gtk_popover_present()
implementation and it shouldn't be necessary anymore. It only breaks
sizing making it grow significantly anyway.
|
|
|
|
|
| |
This shouldn't be necesary anymore and gets in the way now that present
should basically be doing the right thing in GTK.
|
|
|
|
|
|
| |
If we use a layout manager, we will never get a size_allocate call which
we are expecting to setup various assistant positionings. Since we have
a simple bin setup, we can just do the size_allocate and measure manually.
|
|
|
|
|
| |
If code paths are resetting a cell which is essentially already empty,
then we want to short-circuit and do nothing.
|
| |
|