GitLab

When asking a preview from .rst file, we walk back in the file tree
to search for a conf.py sphinx config file.
(the limits are ten level up or the project tree basedir)

the results of sphinx-build commands are keeped around
for each basedir in the tmp folder in gnome-builder-sphinx-build-XXXXXX
like folders (so we can update it after each key strokes or
triggering another preview with the same basedir)

When pasting code with diffrent indentation style then the one in the
Indentation settings, the indentation style of the inserted code stays the
same.

A new plugin retab has been added. The retab is accessible through the
Selection in the popup menu and converts the indentation style base on the
Indentation setting of the file.

https://bugzilla.gnome.org/show_bug.cgi?id=766266

If we get an error and there is no discovered "current directory", we can
just use the builddir as the relative path to the file. This fixes ninja
output from meson for vala warnings.

https://bugzilla.gnome.org/show_bug.cgi?id=780467

This property is used to notify the pipeline that it needs to look at
stdout to extract errors using the error regexes. Typical build systems
such as make, will pass stderr on to the caller. However, others like
Ninja coalesce everything into a single stream, complicating the matter.

This means we burn extra overhead parsing stdout, but in practice, the
output we get from ninja is small enough that it should be roughly
equivalent.

We can reduce how much work we do here by first checking which stream we
are logging to. No need to do regex checks on stdout, and no need to do
directory checks on stderr.

https://bugzilla.gnome.org/show_bug.cgi?id=780467

If we are building out of tree, we will get a path that is based on the
build directory. Instead, we need a path that is relative to the project
root. This will simply trim off the builddir prefix from the path to get
that effect.

https://bugzilla.gnome.org/show_bug.cgi?id=780467

We no longer get the newlines from the build log, so we need not check
for the trailing \n in the logging output.

https://bugzilla.gnome.org/show_bug.cgi?id=780467

We pretty much always want reduced output when building, because things
like flatpak will want verbose for C-I, but that is not appropriate for
inside of Builder.

We need to free the result of g_strjoin.

https://bugzilla.gnome.org/show_bug.cgi?id=780500

This sets the working directory of the launchers (for make, make clean,
etc.) to the directory containing the Makefile, which is important in
case it's a subdirectory of the root project directory (the one with
.git/).

https://bugzilla.gnome.org/show_bug.cgi?id=780499

Honor the PATH environment variable when looking for JHBuild's
executable. Keeping ~/.local/bin/jhbuild as a fallback is useful for
systems where ~/.local/bin/ is not in the PATH by default.

Changes in v2:
- Check for JHBuild's path only once when a project is opened.

Changes in v3:
- Use super() to call the parent __init__.
- Some style nits while at it.

https://bugzilla.gnome.org/show_bug.cgi?id=780373Signed-off-by: Dor Askayo <dor.askayo@gmail.com>

Without them, flatpak and autotools plugins failed to load.

https://bugzilla.gnome.org/show_bug.cgi?id=780503

We should be more vigilant in getting a proper FD number to map in.

We might need access to additional drives.

Keep this lower than the UI priority.

Some git operations such as "git revert" cause the following sequence of
GFileMonitor events: DELETED, CREATED, CHANGED, CHANGES_DONE_HINT. This
means that we need to keep the monitor for a "deleted" manifest in order
to get the following events and recreate it. Unfortunately this also
means we're parsing the manifest and creating a config for it twice for
one set of changes, but not much can be done about that.

When you save an edited file in Builder it renames a temporary file to
the appropriate file name, so we need to deal with that behavior
appropriately for the GFileMonitor that watches the flatpak manifest.
This commit fixes a bug where we were monitoring the temporary file
rather than the actual manifest after such a rename event.

https://bugzilla.gnome.org/show_bug.cgi?id=780237

This commit passes the versioning option specified on the
command line along to the project template so it knows git
is enabled. Some of the templates have some behavior
that's conditional on whether git is enabled.

https://bugzilla.gnome.org/show_bug.cgi?id=780420

If the plugin ships a .gresources file alongside the plugin, we can load
that before loading the plugin. This simplifies plugins to be able to use
GResources even when using a language such as Python.

This does more than just menus, so the name should reflect that.

https://bugzilla.gnome.org/show_bug.cgi?id=780414