Chain up here instead of doing our funky only chain-up after setting
phase. This should make things more predictable for subclasses that
need to do anything funny, and Vala.

https://bugzilla.gnome.org/show_bug.cgi?id=779891

We don't need to require IdeObject for a bunch of things, we can just rely
on GObject + the context property.

This will let us work around some language breaking checks in Vala.

https://bugzilla.gnome.org/show_bug.cgi?id=779891

This reverts commit 5c05725f.

We don't need to require IdeObject for a bunch of things, we can just rely
on GObject + the context property.

This will let us work around some language breaking checks in Vala.

Tasks were not getting completed when they were successful transfers due
to only returning an error and not success.

This is a new widget that can be pushed onto a workbench so that toplevel
messages are visible to the user. You might use this to show a warning
that a package needs to be installed, such as flatpak-builder.

This is a new transfer that can perform installations of packages using
the pkcon command on the host. Such is useful for installing packages
like "flatpak-builder".

Krzesimir Nowak authored Mar 09, 2017 and Christian Hergert committed Mar 09, 2017

It was aborting gnome-shell, when downloading the flatpak failed.

https://bugzilla.gnome.org/show_bug.cgi?id=779802

The first time we show the panel, it is fine to raise it (without
focusing).

We want to default to online docs as they will be the most up to date.
If we don't have that, we'll load from docdir (or translate to the host
if we are accessing via flatpak).

We might need to find the path on the host, so add a helper to join paths
to get access to the file on the host.

If we are not advancing to or past IDE_BUILD_PHASE_BUILD, then there is no
need to clear the previous diagnostics.

https://bugzilla.gnome.org/show_bug.cgi?id=779526

This let's the listener know what stage we are advancing the pipeline to.

https://bugzilla.gnome.org/show_bug.cgi?id=779526

We now rely on the help code that is going into doc/. We can just redirect
to builder.readthedocs.io until we statically generate docs and bundle
them.

We don't really need this anymore, and it just wastes space. We should
instead make our build messages very clear as they come from the
particular build stages/phases.

If a stage is disabled, it will be skipped over. It can be convenient if
we need to disable something except for certain uses.

We want to ensure that all our objects are alive during the process of
the save operation, so ensure that we hold on to them until the last
provider has been saved.

This does change things to happen in a serial (but async) fashion, which
should be fine.

Just some cleanup to use g_steal_pointer() when appropriate and ensure
we set source tags on the GTasks for debugging.

In the past I've seen some weird type issues here, and they seem simple
enough to try to mitigate here.

If we had a really long async process it is possible that our text marks
were destroyed and zeroed. Check for that before calling move_mark.

Now that we can have additional FDs for passing to the host process, we
need to update our assertion for the FDList.

https://bugzilla.gnome.org/show_bug.cgi?id=779494

This allows you to either stop the build (duplicate of the top bar) or
clear the build log from the build log panel. We might also consider adding
a save action here.

https://bugzilla.gnome.org/show_bug.cgi?id=779425

Like one does.

This can end up happening when plugins are enabled/disabled, so we need
to support it. It just means that the user might have to perform a
rebuild if we have something in flight.

Anything to try to be less tricky to various build system tooling.

This reverts commit 5119aa3f.

We don't need this anymore now that we have cursor name.

We won't have access to the theme from the host, so we need to ensure our
settings match what we have available.

By reversing our application of color parsing, we can remove the potential
for uninitialized GdkRGBA.

When the IdeUnsavedFiles starts up, reap old buffers (more than a day old)
so that we don't allow things to fill up ~/.cache/.

We need a location for temporary files that allows us to access them from
the host system (in the case we are moving between flatpak to host).

https://bugzilla.gnome.org/show_bug.cgi?id=779335

Giovanni Campagna authored Feb 28, 2017 and Christian Hergert committed Feb 28, 2017

The gnome code assistance plugin creates an IdeDiagnostic object
with no location, and then sets one or more IdeSourceRange's on it.

We can assume that the first source range corresponds to where the
problem is, and set the location based on it. Without a location,
the diagnostic will be dropped on the floor, which is not nice.

https://bugzilla.gnome.org/show_bug.cgi?id=779359

This was missing from our abstraction, so this wraps the GSubprocess
version and provides a minimal equivalent for IdeBreakoutSubprocess.