| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
| |
Translators sometimes leave extra whitespace, so if we're parsing without
AS_NODE_FROM_XML_FLAG_LITERAL_TEXT then just swallow the trailing spaces.
|
|
|
|
|
| |
Adding this to the metadata means we do not have to look up the metadata source
using the package manager.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
This is used all over the place both in and out of tree.
|
| |
|
| |
|
|
|
|
|
|
| |
Applications have to specify the kudos upstream and relying on these hacky
heuristics only worked some of the time for packages, and none of the time
for xdg-app and Limba.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
For package types without a native klass->compare fall back to something
sensible.
Also add some self tests for this to make sure this never happens again.
Fixes https://github.com/hughsie/appstream-glib/issues/72
|
|
|
|
|
| |
In doing so, we found a number of invalid AppData files, so fix up the values
the best we can.
|
|
|
|
|
|
|
| |
GCab has changed the checksum generation code a few times now, and it's a pain
to deal with. Just use a wildcard to prevent test failures.
Fixes: https://github.com/hughsie/appstream-glib/issues/64
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If we're comparing versions that are both converted to strings it doesn't
really matter how the strings are stored (e.g. AA.BB.CC.DD or AA.BB.CCDD) as
the integer rules still apply **if** we don't remove the prefixed zeros.
The only time we have to be careful is where the firmware metainfo uses the
triplet and the .inf file uses the hex version format.
By using the triplet-style by default we match what Intel and Microsoft are
providing in the firmware release notes.
Basically:
* If you're not using the MS-style versioning scheme, just use hex numbers in
both places.
* If you don't care about applying the update on Microsoft Windows, just
don't supply an .inf file at all.
Fixes the other half of https://github.com/hughsie/fwupd/issues/34
This does break API, but we've never had the method in a released version so
I think it's fine to break at this stage. Make sure you update fwupd at the same
time if you're using git master.
The last bit of this fix would to be somehow encode the "display version" onto
the AppStream <release> object, but this needs discussion upstream first.
|
| |
|
|
|
|
| |
Fixes half of https://github.com/hughsie/fwupd/issues/34
|
|
|
|
| |
There's no point having two implementations of the same thing.
|
| |
|
| |
|
|
|
|
|
|
| |
This changes the default processing element from the desktop file to an AppData
(or MetaInfo) file, and the other sources now refine the base object rather
than the other way around. This matches what Debian is doing.
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
glib's g_malloc and friends now always use malloc internally, as of glib
2.46: https://bugzilla.gnome.org/show_bug.cgi?id=751592
This means we no longer have to reallocate glibc malloc'd buffers and
can expect that g_free works fine on them.
|
|
|
|
| |
... and bump glib dep to 2.45.8 for g_autoptr(GString) support.
|
|
|
|
|
|
|
| |
We still have a few types that don't support g_autoptr(), but very very
close to having everything ported over.
This also bumps gdk-pixbuf dep to 2.31.5 for the cleanup macro support.
|
|
|
|
|
|
|
|
|
| |
realpath() in POSIX.1-2008 supports NULL for the second argument,
returning a newly allocated buffer. glibc has been supporting this mode
for years already.
Thus, switch to this mode and avoid PATH_MAX, which is optional in POSIX
and not provided on Hurd.
|
|
|
|
|
|
| |
This allows us to have an ID that isn't tied to the GUID, and also allows us to
have one component that targets more than one ESRT GUID. This requires changing
the matching logic better, but should match what makes it into AppStream 0.9
|
|
|
|
|
|
| |
This allows the release engineer to _only_ update the metainfo.xml file for
each release. If the firmware.inf file is not being used in Windows for driver
installation then we should make the file static.
|
|
|
|
|
|
|
|
| |
Also, partially revert the 'add filename' commit and move the filename to the
checksum. This allows us to store the checksum of the firmware file as well
as well as the .cab container.
This allows us to validate the firmware on devices using fwupd.
|
|
|
|
|
| |
We need to set this from the build system when doing running the tests with
build systems that do not build in the source tree.
|
| |
|
|
|
|
| |
This makes little sense for other component types such as addons or firmware.
|
|
|
|
|
|
|
|
|
|
| |
This allows us to save the source package or bundle that is referenced to the
release. This allows us to download the correct file when installing a specific
version of a component.
This also allows us to mirror the correct file if the source filename is
different from the location-specified filename, or if there is no location
tag in the firmware metainfo file.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I misunderstood the role of the ClassGuid for UEFI firmware. It is not supposed
to be the ESRT GUID, but instead a generic 'firmware' GUID set by Microsoft.
The actual ESRT is set as a FirmwareID registry key only, so learn how to parse
the HK section and fix up all the unit tests.
As there are .cab files in the wild (cough, ColorHug, cough) that set the
device ID as the ClassGuid, fall back to this to maintain compatibility with
either method.
Many thanks to Mario Limonciello for helping to debug this.
See https://msdn.microsoft.com/en-us/library/windows/hardware/ff547502%28v=vs.85%29.aspx
for more information.
|
| |
|
|
|
|
| |
This allows us to use it with incorporate if requried.
|
|
|
|
|
| |
Using the application-id means that multiple fonts packaged up in the same file
accidentally use the cached image of the first font.
|
|
|
|
|
| |
This removes the panel, which does not work well when the console is not a real
onsole, which is typical in most cases where the builder is being run.
|
|
|
|
|
| |
Distros currently relying on '--screenshot-uri' should start using the
'appstream-util mirror-screenshots' command after the builder has completed.
|
|
|
|
| |
It's no longer used in the status html page.
|
|
|
|
|
| |
There is no need to merge these into one file as we can just install the other
AppStream files manually.
|
|
|
|
|
| |
The gcab firmware files generated in F22 changed as a gcab bug was fixed:
https://git.gnome.org/browse/gcab/commit/?id=b0eb277ab3a1bab9f4408a3cdd0f48819f303968
|
|
|
|
|
| |
This is the first step to decouple the mirroring of the screenshots from the
appstream metadata generation.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If we can't access the network we can't copy the screenshot onto a mirror or do
any of the resizing and padding that we normally do. This is also bad as there's
a slight privacy risk letting upstream know that the user is looking at some
specific software, and typically screenshots are not hosted on https:// so there
may be someone playing MITM games.
Also, but not resizing the screenshot we risk downloading the full size source
image just for the 112px thumbnail. But hey, it's better than nothing.
Resolves: https://github.com/hughsie/appstream-glib/issues/43
|