Mathieu Bridon authored May 27, 2019 and Kalev Lember committed May 29, 2019

This lets people know, before they submit, what data Software will
send, why it is necessary, and who is responsible.

This means we show the same UI if the user installs all the firmware updates or
just one of them. We don't want the user to think the update was done live.

Robert Ancell authored Mar 27, 2019 and Robert Ancell committed May 19, 2019

If an io.snapcraft.* AppStream ID is given, extract the snap name from it.

Although this belies a metadata generation bug, fall back to showing the
generic application icon rather than having a 'hole' in the UI where the icon
would normally be.

This matches commit c955d3a1 that fixed it for the app row class; this
commit does the matching change for the details page.

Fixes GNOME/gnome-software#90

Fixes: GNOME/gnome-software#499

Copied verbatim from PackageKit dnf backend.

Change the variable naming to fit with the rest of the convention across
the code base. See
GNOME/gnome-software!212 (comment 512566)

.

This makes no functional changes.
Signed-off-by: Philip Withnall <withnall@endlessm.com>

Change the variable naming to fit with the rest of the convention across
the code base. See
GNOME/gnome-software!212 (comment 512566)

.

This makes no functional changes.
Signed-off-by: Philip Withnall <withnall@endlessm.com>

Change the variable naming to fit with the rest of the convention across
the code base. See
GNOME/gnome-software!212 (comment 512566)

.

This makes no functional changes.
Signed-off-by: Philip Withnall <withnall@endlessm.com>

As with commit bc855b24

. Some of these are existing issues, and some
are potential issues which would surface when adding to the code.
Signed-off-by: Philip Withnall <withnall@endlessm.com>

It’s possible for flatpak_installation_fetch_remote_metadata_sync() to
fail — if so, g_bytes_get_data(NULL) will be called, which will emit a
critical warning.

Correctly handle the error from libflatpak.
Signed-off-by: Philip Withnall <withnall@endlessm.com>

If a GError defined in a wider scope is used in a loop, even if defined
using `g_autoptr()`, it needs to be freed before the next loop
iteration.
Signed-off-by: Philip Withnall <withnall@endlessm.com>

Instead of always leaving out the first subcategory, be a bit more
careful and do an ID comparison to skip the correct category we want to
skip.

This fixes Addon::Font to correctly show up again in the category view
(regression from commit d3648252).

Fixes: GNOME/gnome-software#133 (comment 509825)

This is the same as the previous commit, just for flatpak. To avoid
deadlocks, this commit also adds _unlocked() versions of two public
functions for internal use.

We would access priv->silo in one thread and the do
"g_clear_object (&priv->silo);" in another, invalidating the silo that
the other thread was using, leading to crashes as the thread that was
using priv->silo would access a destroyed XbSilo at that point.

As we clear the XbSilo instance and build a new one, the locking has to
be done inside gnome-software code and can't be fixed inside XbSilo.

This commit adds locking using GRWLock, so that we only take the writer
lock when rebuilding silo, and otherwise take the reader lock when we
just query the silo.

... so that we can use this on older glib that don't have glib MR
GNOME/glib!825 .

This was alone leaking 1.5Mb of RSS every time we refreshed.

Just use 'rpm-ostree' as origin to get something that's better than an
empty field. We'll need to implement /etc/ostree/remotes.d parsing in
the future, but this should do for now.

The new GsPackageKitHelper class introduced in mega-commit 37103e6e only
handled progress marshalling when downloading a list of packages, but
not when the PK transaction is one big monolithic operation that only
has a single GsApp associated, such as when downloading all of the
distro upgrades, or when refreshing the cache.

This commit adds a way to explicitly add a GsApp to GsPackageKitHelper
to use for marshalling progress in these cases.

Fixes: GNOME/gnome-software#620

Be more careful to avoid dereferencing priv variables in an async
callback handler, which can be called when the dialog is in dispose and
the priv variables have all already been cleared to NULL.

Note that this needs fixing in many more places, but this particular
crash is a top F30 crasher.

https://bugzilla.redhat.com/show_bug.cgi?id=1634075

The plan for Fedora Silverblue is to use rpm-ostree layering for things
that are non-apps (e.g. hardware support, codecs, fonts etc that
supplement the base system), and only offer apps through flatpak.

This commit makes it so that available apps don't show up in search, but
anything already installed still gets correctly shown (so that they are
removable / launchable /updatable etc).

Kalev Lember authored Apr 19, 2019 and Richard Hughes committed May 02, 2019

This allows plugins to control if apps are discoverable or not.

Discoverable here means that they get hidden from search results and
category views, unless they are already installed. This makes it
possible to launch and remove and update already installed apps, but
hides them otherwise.

This is going to be used in the rpm-ostree plugin to hide package-based
desktop apps on Fedora Silverblue and only show flatpak apps by default.

Philip Withnall authored Apr 30, 2019 and Richard Hughes committed May 02, 2019

This has been upstreamed from Endless’ fork, where it was originally
added as a paid-for translation. I can’t speak Marathi, so can’t vouch
for the quality of the translation.

Only one string is translated, but that’s probably better than nothing.
Signed-off-by: Philip Withnall <withnall@endlessm.com>