This uses a menu and is thus subject to the menu icons visibility
setting, but here it should reflect the view from the symbols list,
and thus show the icon in all cases.
Don't use the files inode as the hash. Although it looks like a good
idea for de-duplicating links as well, it has several issues, including
non-uniqueness of inodes across file systems.
The way it was done hashing the inode but comparing the file name
string pointers also made the hash mostly irrelevant, as it just stored
filenames sharing the same inode in the same hash bucket but without
actually doing any de-duplication, making the whole thing a convoluted
way of converting to a list.
Instead, hash and compare the filenames themselves, which, even though
it doesn't handle links de-duplication, is better than the
non-functional previous code.
Also, directly build the list and only use the hash table as a way for
checking for duplicates, which is both faster and gives a stable
output.
Now the keybinding can be overridden (e.g. using Tab for it as well as
normal behavior), beeping when there is no next cursor is more annoying
than useful.
Part of #1554.
The new styles are for "diff of a diff", e.g. lines starting with `++`,
`+-`, `-+` and `--`. Those are currently mapped conservatively keeping
the previous display behavior, until we have good style matches for
them.
Scintilla 3.7.6/4.0.0 deprecated `SCE_*STYLEBITS*` and moved it to
deprecated features that require a build-time flag to be available.
Thus, drop use of those (as they are now no-ops anyway) and bump the
ABI (so plugins depending on those don't build mistakenly load) and API
(so a developer can guard use of those if wanted) version accordingly.
Our custom scroll handler for horizontal (Shift+Scroll) and page
(Alt+Scroll) scroll didn't properly check the scroll direction and
assume that if it's not down it's up. This was mostly not a problem
because the other types only were left and right scroll events, which
are a lot less common.
However, it became a lot more problematic with GTK 3.4 that introduced
"smooth scrolling", and thus a new scroll type that can happen for
events in any direction. We then would scroll up (as we assume "not
down" is up) regardless of the actual direction of the event.
It's still not clear why we'd get smooth scroll events on X11 as no
code I can find asks for it and we generally don't get those, but
sometimes a Scintilla widget starts receiving them, leading to the bug.
On Wayland on the other hand, Scintilla asks for smooth scroll events,
so we need to have a fix for it in any case.
Make the brief text be distinct between msgwin_*_add and msgwin_*_add_string().
Also add @see directives where appropriate. Lastly, add @since to
msgwin_status_add() for completeness.
The variadic variants cannot be gobject-introspected, i.e. are not available
in Peasy.
In fact, msgwin_compiler_add_string() and msgwin_msg_add_string() already
existed and have just been exported. msgwin_status_add_string() is new but
msgwin_status_add() becaume a wrapper around it in the same fashion as the
other two pairs.
This prevents loading a spurious tag for the format specifier line, as
well as fixing loading of CTags format with a format specifier line.
Before this change, the file pointer was rewound after reading a format
specifier line; but this had various unwanted side effects depending on
the recognized format:
* For TagManager and Pipe formats, it led to loading a tag named after
the format specifier (e.g. a literal "# format=tagmanager"). This
was fairly harmless and only introduced a spurious tag seldom even
used because "#" isn't usually considered for looking up completions.
* For CTags format, having an explicit specifier led to failure to load
the file in most cases because the specifier line would be parsed but
doesn't usually follow the format's requirements, leading to early
abortion loading that file. On some very specific specifier lines
actually following CTags format, it could have led to loading a
spurious tag instead.
Fixes#1814 and closes#1816.
When toggling a plugin, we temporarily set the tree store's row entry
for the plugin pointer to NULL as we destroy and reload the selected
plugin, and its pointer would be invalid in the meantime. This results
in the filter we use to display search results to temporarily hide the
row, changing the actual number of rows and thus, depending on timing,
this will or will not change the selected row (it will when double
clicking, not when single-clicking), in a seemingly more or less random
fashion as we use a sorted model.
Finally, as we manually update the buttons visibility for the toggled
plugin (as we otherwise do only for changing selection, which should
not happen in this case -- well, most of the time as you can see), this
can lead to the buttons to be updated for a now unselected row, getting
those out of sync.
The fix here is not to actually hide rows with a NULL plugin, because
it can only happen in 2 cases, where we actually want to see it:
1. while toggling a plugin, as explained above, in which case it had to
match the search already.
2. when there is no plugins and we want to display a "No plugins
available" message, and the search should not affect this.
This incidentally also fix the "No plugins available" so it's actually
visible, instead of always hidden.
Fixes#1781.
This was introduced in 1f71ccd because the VTE widget's existing
button-press handler didn't return `TRUE` and so the event propogated
up to the notebook showing both of their context menus.
For some reason the TravisCI build has `-Waggregate-return` enabled as
an error even though it's completely valid code. Also constify the
input `color_name` argument to `load_color` function.
* Improve GTK+3 theme loading
- Merge geany-3.0.css into geany.css
- Load geany-3.20.css to override if runtime version dictates
- Load geany.css from app->configdir if it exists.
* Add initial documentation for GTK+ CSS theming
* Add geany.css file to the `Tools->Configuration Files` menu
Also mark such config files as changed if they don't already exist to
force the user to save or discard any newly created config files.
* Allow customizing message window using GTK+ themes
* Improve GTK+3 CSS docs a bit
State which classes are meant to be overridden by themes.
* Go back to conditionally loading geany-3.0.css
It can't be merged into geany.css since GTK+ CSS parser doesn't ignore
invalid properties when prefixed with vendor extension (ie. -Gtk).
* Don't mark unedited config files as changed
They probably should be marked as changed since they don't exist on
disk yet, but for the time being leave it how it was.
* Fix misuse of CSS ID vs class in manual
* Minor colour value tweaks
Based on feedback from @b4n
To document when switch cases fall through intentionally, hiding the
warning by GCC at the same time.
Moved the break statement outside of the if conditional in keybindings.c
as I believe this was an actual bug found by GCC's warnings.
