The call to g_proxy_resolver_get_default() is (transfer none) and so unref'ing
it causes GLib to get very upset. Also, move it to be per-FwupdClient rather
than per-request as it can be expensive to call in the PAC-runner case.
Only a small amount of the firmware on the LVFS will be available. The user
can use --ipfs on the command line for testing, or change the system-wide
default in /etc/fwupd/daemon.conf.
The IPFS daemon and command line client will need to be installed manually.
We forgot to include FwupdClientDownloadFlags when adding the original method
fwupd_client_install_release() -- and we want to use additional download flags
for operations in the future.
The metadata might want to pass more than one location URI to the client, for
instance if the file is available from more than one HTTP mirror.
Use the noun of location to match the AppStream <artifact> naming; this is the
last place where LVFS AppStream diverges from the official specification and
it would be good to bring fwupd back into line -- although the LVFS will have
to write both elements for a very long time.
See https://www.freedesktop.org/software/appstream/docs/chap-Metadata.html
Also: we're not changing the format of the `Uri` GVariant key to preserve both
forward and backwards compatibility of the library. We can remove it when we
next break API.
We don't want the (shared) progress notifications going to the thread-default
GMainContext by default as this changes behaviour (ABI?) with older libfwupd
versions.
If the client really does want progress notifications to go to a different
GMainContext they it can call fwupd_client_set_main_context() with the desired
context, e.g. g_main_context_get_thread_default() or g_main_context_new().
This has the intended side effect of returning progress and state notifications
to the main thread, which is typically (but not always) the thread that created
the FwupdClient object.
When using the sync API entrypoints like fwupd_client_get_devices() these call
into fwupd_client_connect() to ensure the proxy is set up. This is not thread
safe and chaos ensues if you call two sync functions from different threads,
like GNOME Software does at startup...
Use a mutex to protect access to this shared resource.
Original test code from Philip Withnall, slightly tweaked by me. Thanks!
The end year is legally and functionally redundant, and more importantly causes
cherry-pick conflicts when trying to maintain old branches. Use git for history.
For fuzzing we want to exclude libcurl support as it depends on other very heavy
libraries like OpenSSL or libtasn which make the fuzzing binary much larger if
linked statically.
Some vendors really really want 1.5.x in newer RHEL versions, but the version
of curl is too old. Add #ifdefs so that we can emulate (somewhat imperfectly)
the 'new' CURLU functonality.
Use the 'default' main context by default so that we still get the signals
delivered via g_object_notify() but allow the library user to specify an
alternate GMainContext as required.
Using g_main_context_push_thread_default() works for PackageKit as signals are
returned using g_signal_emit() which is synchronous, and so doesn't use the
main context at all.
Use the same style of API which we did for libgusb all those years ago instead.
This fixes a regression from 2f49da7f4e where we
tried to parse the GPG .asc file as a JcatFile even although the remote had
Keyring=gpg (which is the default for ODM accounts on the LVFS) which returned
the cryptic 'Invalid compressed data' message.
The workaround for 1.5.2 is to change the local remote.conf from 'Keyring=gpg'
to 'Keyring=jcat' and to save the file.
The former drags on glib-networking and then gsettings-desktop-schemas, which
add over 5Mb to the minimal IoT and CoreOS composes. Everything already uses
libcurl (even NetworkManager!) and so this is an easy way to reduce image size.
For instance, we can tell the user that UEFI UpdateCapsule is disabled in the
system firmware, or that efivarfs is not mounted. This is much better than
creating "dummy" devices which are really just hacks around the problem because
no better API existed. THe dummy devices cause as many problems as they solve.
Plugins have to set FWUPD_PLUGIN_FLAG_USER_WARNING if a warning should be shown
to the user, and only one warning will be shown of each failure type.
It is expected that GUI clients like gnome-software and gnome-firmware would use
this API to notify the user the localized message for why firmware updates are
not being shown.
Fixes https://github.com/fwupd/fwupd/issues/2456
The FWUPD_INSTALL_FLAG_FORCE flag has really unclear semantics, and ignoring a
file CRC, checksum or model ID should only be done when using fwupdtool actually
debugging a plugin or firmware parser.
Use the existing --force flag when we want a "gentle nudge" like reuploading
previously processed reports.
This changes allows for downloading firmware from a remote server
pointed from a local remote manifest.xml.gz file
Change-Id: Id00870f9c2817d48d6d301d2b6d229ba1ca6045a
This is required when the calling application needs the low-level soup-session
with the user agent set correctly rather than using the helper methods like
fwupd_client_download_bytes().
This is what GNOME Software needs to handle the GsApp progress completion.
This takes care of downloading the correct files and allows remotes to be
refreshed from other CLI and GUI tools without copying large chunks of code.
This also allows us to download the metadata without writing two temp files
to the users cache directory. Although not security sensitive, it's probably
not a good idea if we can avoid it.
