| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
This way, things will continue to build on Visual Studio.
|
| |
|
|
|
|
| |
This is the test case we needed from #190 to actually verify a bug fix.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When porting to PCRE2, we only implemented the syntax highlighting engine
because there was a lot to port. That meant we used PCRE for search
contexts (via GRegex), but PCRE2 for highlighting.
Now that we have enough implementation in our PCRE2 wrapper, we can make
use of ImplRegex for search contexts too.
It's possible we have some fallout, but for what I've seen in the wild
thus far, the bases seem to be covered.
It's unlikely that we'll backport this to GtkSourceView 3.x or 4.x series.
However, if there is enough interest for someone to do that work and
really test it on platforms in the wild, we'd be willing to manage that
upstream with them.
Fixes #138
|
|
|
|
| |
We need additional work to make things match exactly with JIT enabled.
|
|
|
|
|
| |
The goal here is to be able to use PCRE2 from search-context, so we want
to test that usage to find any issues.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
This is a barebones test to make sure that we can load some of our files
from the tests/syntax-highlighting/ directory. Ideally, we would process
more of this so we can find errors with PCRE2 loading, but this is a first
step in that direction.
|
|
|
|
|
| |
This doesn't hold up everywhere depending on the content engine
behind things in GLib. So we can't guarantee this on Mac.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
These should be "const gchar * const *" since we're breakin API/ABI
anyway. This is not an ABI break, but it will cause warnings in existing
code that was passing "gchar**". This is more correct however since we
do not take any references or mutate the strings, or pointer to strings.
|
|
|
|
|
| |
These are no longer necessary per updates to HACKING. So remove them to
be consistent with our suggested contributions.
|
|
|
|
|
| |
Most notably, using gtk_init(), adapting to <layout> changes in GTK 4
Builder XML, and renderers becoming widgets.
|
|
|
|
|
|
|
|
| |
Undo is now managed by GTK itself (with a more memory compact form of
the undo manager from GtkSourceView). It has the added benefit that it
is run from the default GtkTextBuffer handlers right next to the code
that mutates the B-Tree, meaning that we have a better chance at avoiding
reply/reordering issues which can happen today.
|
| |
|
|
|
|
|
|
| |
- Use G_DECLARE_
- Remove use of private instance
- Move to -private.h
|
|
|
|
|
|
|
| |
- Use G_DECLARE_
- Move to -private.h
- Make final
- Fix multi-line function indentation
|
|
|
|
| |
Keep G_GNUC_INTERNAL things inside of -private.h headers.
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
- Use G_DECLARE_
- Make final
- Use private header
|
|
|
|
|
| |
This can get dropped when using the static linking, so we need to ensure
we compile it in instead.
|
|
|
|
|
| |
This will allow us to use more aggressive optimizations on the primary
libgtksourceview-4.so such as -Wl,Bsymbolic and -fno-plt.
|
|
|
|
|
| |
We always want assertions as part of the testsuite, even when running
release or plain builds as done by distros.
|
| |
|
|
|
|
|
|
|
| |
As we try to make styles more declarative, we need a fallback to ensure
that we don't break existing style-schemes in the wild.
Related to !77
|
|
|
|
| |
Fixes #43
|
|
|
|
|
|
| |
On various beta systems for GNOME 3.34, we might not have access to the
at-bridge. This fixes the test system to work without it, which isn't
really something we want to be specifically testing anyway.
|
|
|
|
|
| |
We have fully transitioned to meson now, so it is time to remove the
venerable autotools.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Unfortunately, Visual Studio builds do not really like the concept of
linking 2 static libraries and not using any sources for building a
given DLL (it will not include those items when linking the final DLL as
it will optimize most of the items out as the linker does not believe
those items are being really used in the final DLL), so we need to do
the following for the Visual Studio builds:
-Define a separate GtkSourceView shared library target that contain the
core sources *and* the word completion provider sources, along with the
generated glib-mkenums and GResource sources. This is actually how the
Visual Studio projects build the sources for the GtkSourceView DLL, due
to the same reasoning.
-Declare the various internal dependencies accordingly; note that we
still need the static core library for building the test programs.
This means that on Visual Studio builds we need to build the core
sources 2 times, it seems that this is unavoidable at this point.
The build files for the tests in tests/ and testsuite/ has been updated
accordingly so that things will build and link. Note that for the
test-stylescheme test, we need to ensure that we only link to the
static core library, otherwise it will fail to link.
-Update the enumeration source generating process, so that we include
config.h during the build and so ensure that the symbols in the
enumeration sources are properly exported. This is necessary
especially for the introspection builds on Visual Studio, as well as
for the test programs.
-For builds prior to Visual Studio 2013, we must include our math.h so
that we can have a fallback implementation for round(), which is only
provided in Visual Studio 2013 and later.
|
| |
|
|
|
|
|
|
|
|
| |
This commit provides an initial port to the meson build system. See
meson_options.txt for the changes to command line options. For a full
build of GtkSourceView, the following will work.
meson -Dglade_catalog=true -Dinstall_tests=true -Dgtk_doc=true build/
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In GLib 2.59, the iteration order of GHashTable was changed (its order
has always be, and will always be, undefined). This has broken some
comparisons in check_language() in test-language.c, which assumed a
certain hash table ordering for the results returned by
gtk_source_language_get_*().
Fix that by using an unordered comparison function. This should be
backwards compatible with older GLib versions.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
|
|
|
|
|
|
| |
The previous commit removed the filenames, but didn't remove the line,
since in most cases it wasn't necessary (and the diff --stat was clearer
if it was the same for every file, i.e. only one line modified).
|
|
|
|
|
|
|
|
|
|
|
| |
The LGPL doesn't require it. When renaming a file, we often forgot to
update the filename. And when the namespace will be renamed, it'll be
one less thing to care about.
Done with this command (or a few other variants):
$ sed -i 's/^ \* gtksource.*\.[ch]$/ */' *.[ch]
and then by looking at the diff to see if it didn't remove something
else.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
_gtk_source_iter_get_leading_spaces_end_boundary() moved to the
following lines if the line at iter doesn't contain text, because
g_unichar_isspace() returns TRUE for \n and \r too.
But after a quick look at the code calling
_gtk_source_iter_get_leading_spaces_end_boundary(), the bug didn't have
an impact. It had an impact in Gtef, where I've copied the function and
used it for another purpose.
|