I avoided this for the longest time because it really tanks scrolling
performance. But we can at least have it as an option and disable it
by default.

This should be more similar to the path that Flatpak uses automatically.
It fixes some issues with gdbus-codegen not running properly due to
improper relative path access.

flatpak run org.gnome.Sdk was used to see what the default $PATH should
be and this matches it.

When we do not have an indenter, we try to copy the indentation
style of gedit. We simply copy the previous line. However, we
were copying too much information by not stoping at the line
end.

This caused multiple newline insertions matching up to the
number of new lines after the insertion cursor (plus the newline
we added).

https://bugzilla.gnome.org/show_bug.cgi?id=785213

Now that we mimic GtkSourceView, we do not need to setup the base class
to do it's own indentation.

https://bugzilla.gnome.org/show_bug.cgi?id=785425

Sébastien Lafargue authored Aug 04, 2017 and Christian Hergert committed Aug 29, 2017

The previous commit 75943d28 was wrong but we need
to add modifiers filtering so that e or f with
shift, control or alt are not grabbed by the indenter

Sébastien Lafargue authored Aug 04, 2017 and Christian Hergert committed Aug 29, 2017

https://bugzilla.gnome.org/show_bug.cgi?id=785806

Settings set in our layer take precedence over children.

In some situations, we know we want to indent, but we don't have
a smart indenter. We can instead fallback to a GtkSourceView-
like indenter which copies the previous line.

This is done by using NULL as the @self parameter to the
various IdeIndenter API.

https://bugzilla.gnome.org/show_bug.cgi?id=784633

We can have a very simple indenter as a fallback for the
interface. Just pass NULL for @self in those cases.

This tries to mimic GtkSourceView in that it just copies the
previous lines indentation (when possible). No indentation
is performed if we land on the first line of the buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=784633

To pacify damned-lies.

This also unifies into a single ignore function so that we don't have to
duplicate things.

These can happen in various places where we read an id/branch/arch out of
ancillary data and it is empty. In most places we use g_strcmp0() to be
NULL-safe, to to simplify those code-paths, just handle NULL here.

https://bugzilla.gnome.org/show_bug.cgi?id=782678

Mohammed Sadiq authored May 13, 2017 and Christian Hergert committed May 17, 2017

https://bugzilla.gnome.org/show_bug.cgi?id=782580

When the build configuration management changed to using
IdeConfigurationProviders rather than doing everything in
IdeConfigurationManager, the default configuration stopped persisting to
the disk (so changes made are only effective during a session). This is
because the configuration was being added by the manager as an
IdeConfiguration rather than an IdeBuildconfigConfiguration, and
IdeBuildconfigConfigurationProvider knows how to read and write
".buildconfig" files.

The most obvious solution, creating the default configuration in the
IdeBuildconfigConfigurationProvider's load function, doesn't work because the
loads are asynchronous and there has to be at least one configuration
when the IdeConfigurationManager finishes initializing (otherwise the
IdeBuildPipeline will fail to initialize).

Instead, the load interface for IdeConfigurationProviders was changed to
an async/finish pair, so the IdeConfigurationManager knows when the
loads finish. At that point, it can check if a configuration was
restored from a .buildconfig file (in which case nothing needs to be
done) or if the default configuration was added by the
IdeConfigurationManager (in which case the buildconfig provider needs to
be informed of it so it can be persisted when changes are made).

https://bugzilla.gnome.org/show_bug.cgi?id=779240

If the config-opts line is the last item in the module, it shouldn't end
with a comma. Also, the right curly brace needs to be written before
continuing to the next iteration of the loop. Otherwise the following
line will be written before the closing curly brace.

We need to add a comma to the previous line to keep the JSON valid, but
without using the pointer after it has been freed.

This checks that the marks we got from the source snippet are valid.
Otherwise the textiters could be left in an invalid state.

This also checks to ensure that animations are enabled before calling the
animation helper.

https://bugzilla.gnome.org/show_bug.cgi?id=782701

When performing our operation to activate the clang completion item, we
want to ensure the iter passed to us stays valid. We also want to pass
the iter location into push_snippet just to ensure we get the same
location.

https://bugzilla.gnome.org/show_bug.cgi?id=782701