Promises and by extension async/await is unsuable without this.

See https://gitlab.gnome.org/GNOME/gjs/-/issues/468

This makes the icon-name writable only, and icon/expanded-icon readwrite.

This apparently isn't necessary even though GtkTreeExpander does it.

<Company> you don't need list_row_notify_depth() btw because depth never changes
<Company> the listrowitem will go away before that happens

This should fix an "empty notification title" in the omnibar when building
with not message received from the workers.

Adds some radio rows to preferences to toggle on/off rust-analyzer
features. In the long run, we'll likely need something significantly more
tweakable for this stuff, since there isn't really a way to pass these
objects along to the user to tweak advanced LSP features.

Fixes #1795 by way of enabling the nightly option.

Without this the synchronous check causes Builder to stutter because
'podman exec' adds almost 100ms of an overhead execution time to the
executed command.

Based on d907158e & 6513765c

Fixes https://gitlab.gnome.org/GNOME/gnome-builder/-/issues/1794

We don't need a mutex per-runtime for this, since you only really ever
access one at time and I don't want to deal with the complexity of sharing
them with the launcher mutex.

This just uses plain-ol' G_LOCK() to synchronize cache access between
potentially threaded access of the runtime.

Related #1794

This ensures the project is available in the recent projects at the greeter
if you return to it without exiting Builder.

Fixes #1667

This was missing from the IdeSourceView port, and it is a bit nicer to
keep it out of IdeSourceView anyway.

Long term, it would be nice to have a refactory dialog so that we can
show the upcoming text edits before applying them.

Fixes #1695

To keep CI going.

For CI and all that.

As we don't have a 43 yet to use.

This was also still missing from the GTK 4 port.

This was missing from the GTK 4 port of IdeSourceView still.

We removed the foundry items from the tree, so these aren't useful anymore.

We need to be sure that the GtkListBox has access to it's references so
that they can be removed in-case the model changes. Additionally, we don't
want it to continue to react to changes, so just clear the model. Odd that
it wasn't done this way to begin with.

Fixes #956

It's nice to keep the object around across the signal callbacks.

While I'm looking at things here, port to GSignalGroup and make the
signal group creation/destruction symmetric w/ init/finalize.

Fixes #956

As it can break input methods currently.

This way we can separate items that have names and their display format
(which is pango markup). The only thing using that was the xml symbol
node, which we can quick fixup.

This fixes showing things like impl<foo> in the symbol tree from languages
that, you know, aren't xml.

This uses the formatter for the buffer to format the document upon save
if the format-on-save GSetting is set. It requires the setting to be
enabled and an IdeFormatter registered for the document syntax.

Additionally, since the buffer does not have access to view settings like
the size to draw indentation, we rely on IdeFileSettings to get those
values. In general, those should be in-sync with the document view since
it uses those to bridge to the display too.

Related #1795

This was missed during the port, but the goal here is to elevate some
messages about the decoding process back to the application. It still
needs some fixes on the sysprof-agent side though.

We might not get a value back from the peer for this, so assume that
period is always going to be wanted unless the peer gives us a character
set to work with.

Related #1800

We want to do this very last so that the reveal-start/end/bottom state is
always going to be right, even if other panel restoration raises the
widget (thus showing the panel).

Having something is better than nothing, so try to fill these out a bit
more until we can create icons for them.

Now that we have top-match to elevate perfect matches (like snippets) we
can lower the snippet provider priority, raising more important providers
to the top.

This is causing issues with the hover provider, so the safest fix for the
time being is to disable selection.

Related #1801

I have no idea why a Fedora 37 VM gets some extra space when this grows,
but this might be a better way to ensure that it ends up right rather than
the deprecated property (which was not deprecated mid-cycle when this code
was written).

Similar to gd/gD in Vim.

This simplifies a lot of the layout for really long strings and helps
ensure that our url bar doesn't grow unweildly wide.

Fixes #1802

We aren't in a good place if we're relying on IdeSubprocessLauncher,
particularly from pipelines as we can get into really weird state with
how env FOO=BAR are processed, sh -c '', and more.

Give us some headway here by providing access to a run context so that
we can incrementally fix these without breaking ABI this cycle.

This helps ensure things are executed correctly without all the madness
of guessing via subprocess launchers.

We don't want people using the create_launcher() path because it really
isn't compatible with how run contexts work. You could end up with
something like sh -c 'env FOO=BAR something' (and now what if you need
to append to that).