summaryrefslogtreecommitdiff
path: root/libappstream-builder/asb-task.c
Commit message (Collapse)AuthorAgeFilesLines
* trivial: Fix some typos spotted using codespellRichard Hughes2019-04-081-1/+1
|
* Use the SPDX-License-Identifier in all source filesRichard Hughes2019-02-091-15/+1
|
* trivial: Show a useful error message if decompressing a package failsRichard Hughes2017-11-031-1/+2
|
* Actually return failure if the AsbTask failsRichard Hughes2017-06-011-58/+38
| | | | Now that we're not threaded we can provide a sane error.
* trivial: Fix a variable with a misleading nameRichard Hughes2017-06-011-7/+7
|
* Remove the cache-id functionalityRichard Hughes2017-05-161-24/+0
| | | | | | | | | The results when using the partial update were not always the same as doing a fresh run (some components depend on other components) and even I'm not using this functionality anymore when running appstream-builder on the whole Fedora, achive, even across on all branches. Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1426166
* trivial: Reset GError when skipping a fileKalev Lember2017-01-091-0/+1
|
* Limit recursive deps lookup to the same source packageKalev Lember2017-01-091-2/+8
| | | | | Otherwise we may end up inadvertently marking an app as ModernToolkit when all it does is require zenity.
* Revert "trivial: Prevent recusion when a package lists itself as a dep"Kalev Lember2017-01-091-4/+0
| | | | | | | Commit 49749de0bd597e5e9fe5e4b92875d6730a7b6968 fixed this in a more general way. This reverts commit b04fb193dad0e6e977d9bda2487bd50075f8f61c.
* Revert "trivial: Do not recurse for library or filename deps"Kalev Lember2017-01-091-8/+0
| | | | | | | This led to ModernToolkit kudos being skipped for apps where a subpackage links with a modern toolkit library (gtk+ 3 or Qt 5). This reverts commit e7f14579d1f9ab6f0f016caf81c34bb576c9ea92.
* trivial: Do not recurse for library or filename depsRichard Hughes2017-01-041-0/+8
|
* trivial: Do not follow a foo->bar->foo->bar depchainRichard Hughes2017-01-041-1/+3
|
* trivial: Prevent recusion when a package lists itself as a depRichard Hughes2017-01-041-0/+4
|
* Recursively add subpackage deps into the main packageKalev Lember2017-01-041-10/+86
| | | | | | | | This fixes appstream data generation for packages that have long dep chains and icons are at a deeper level than in the immediate child package. e.g. in Fedora, 'marble' package depends on 'marble-widget-qt5' which depends on 'marble-astro' which depends on 'marble-common' where we can finally find the app's icon.
* trivial: Remove a libreoffice hack that's no longer neededKalev Lember2017-01-041-4/+0
|
* Don't add <releases> for addonsRichard Hughes2016-12-061-3/+5
|
* Add two new GCC warnings and fix up signed/unsigned issuesRichard Hughes2016-07-251-1/+1
|
* trivial: Get rid of useless internal-only gtk-doc markupRichard Hughes2016-06-281-18/+0
|
* Add yast2-branding-openSUSE if an app depends on yast2Dominique Leuenberger2016-05-191-0/+9
| | | | | Applications depending on yast2 have an implicit dependency on yast2-branding-openSUSE, which brings required icons in this case.
* trivial: Add a work-around for the new libreoffice in the builderRichard Hughes2016-04-291-0/+5
|
* Only use the package license when not provided by the AppData fileRichard Hughes2015-12-091-1/+2
| | | | | In doing so, we found a number of invalid AppData files, so fix up the values the best we can.
* Use AsStoreCab to read .cab files when building metadataRichard Hughes2015-10-071-0/+36
| | | | There's no point having two implementations of the same thing.
* Require AppData files to be present in the AppStream metadataRichard Hughes2015-09-161-12/+0
| | | | | | 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.
* Do not overwrite the AppData URL with the one from the packageRichard Hughes2015-09-161-1/+2
|
* trivial: Drop as-cleanup.h include where not neededKalev Lember2015-09-081-1/+0
|
* Use GLib's cleanup functions instead of libgsystem'sKalev Lember2015-09-081-5/+5
| | | | | | | 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.
* libappstream-builder: Port to G_DECLARE_DERIVABLE_TYPE macroKalev Lember2015-09-081-3/+2
|
* Drop the unused _len arguments on many functionsRichard Hughes2015-08-031-7/+5
| | | | | | Using the length of -1 for 'unknown' breaks the GIR-generated bindings. This breaks API and ABI and the soname has been updated to reflect this.
* Remove the batch output option from the builderRichard Hughes2015-04-071-49/+7
| | | | | 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.
* Speed up adding extra packagesRichard Hughes2015-02-041-4/+6
| | | | Do not load the filelist if we're just checking for a common srpm name
* Fix a builder crash where a package depends on itselfRichard Hughes2015-02-021-0/+2
|
* Also clear the filelist and deplist of the extra packageRichard Hughes2015-02-021-0/+7
| | | | This typically halves the amount of RSS when the context is unloaded.
* Clear the dep list and the file list if we've finished with the packageRichard Hughes2015-02-021-0/+3
| | | | This is a simple way to further decrease the peak RSS value by a few hundred MB.
* Close the packages after they have been processedRichard Hughes2015-02-021-0/+2
| | | | | | | | | This means we have to re-open the files for the few packages used as extra-packages to other packages, but means we keep the RSS value to a sane value for a large number of packages. We can't go above 4GB RSS for Fedora... This also means we spend less time 'shutting down' as the packages have typically been released as part of the task, rather when the context exits.
* Add the extra child deps to the main packageRichard Hughes2015-01-301-19/+32
| | | | | | This means if X depends on X-libs, and X-libs depends on GTK-3, then X should have the modern toolkit kudo. This means switching around the API a bit so that we can add deps after the AsbPackage has been created.
* Make the failed AppStream metadata optionalRichard Hughes2015-01-241-3/+3
| | | | | | | Also, clean up a lot of booleans into one sane bitfield. The libappstream-builder library isn't API stable and has no external users, so we can move things around a bit without worrying.
* trivial: Fix potential builder crash when a package has no depsRichard Hughes2015-01-221-1/+1
|
* Add the VCS information to the AppStream metadataRichard Hughes2014-12-171-0/+8
| | | | Resolves: https://github.com/hughsie/appstream-glib/issues/27
* trivial: Don't show warnings when doing distcheckRichard Hughes2014-11-041-3/+3
| | | | | It seems that the compiler really does require an initializer when using the _cleanup_ macros.
* Save the screenshots as soon as possibleRichard Hughes2014-10-231-0/+8
| | | | | | | We don't want to store +4Gb of pixbuf data for the lifetime of the builder. We can make the assuption that any application shipping screenshots also ships a valid AppData file, and dump the screenshots to disk immediatly after the task has completed. This means we can process Fedora rawhide with <2Gb RAM.
* Write the application XML after all the vetos have been addedRichard Hughes2014-10-201-5/+0
|
* Do not leak applications when building if more than one plugin returns resultsRichard Hughes2014-10-151-2/+8
|
* Only save application resources after the application have been mergedRichard Hughes2014-10-151-21/+0
|
* Only detect missing data after the application have been mergedRichard Hughes2014-10-151-12/+0
|
* Add asb_context_set_embedded_icons()Richard Hughes2014-10-011-0/+8
| | | | | | | This allows us to specify when building the AppStream metadata that we want icons embedded into the output XML. This is going to be slow and lead to a giant XML file when used for whole-archive operations, but makes things simpler for smaller repos or single packages.
* Use AsIcon to represent an abstract iconRichard Hughes2014-10-011-1/+1
| | | | | | | | The assumption that each application had just one icon, of a single type is much too restrictive. We need to support stock fallbacks of missing cached icons and also need to support multiple sizes of embedded icon. This breaks API.
* Only get the package data when requiredRichard Hughes2014-08-281-15/+43
| | | | This speeds up reading a large number of packages.
* Rename as_app_get_id_full() to as_app_get_id()Richard Hughes2014-08-221-2/+2
| | | | There's no more non-full ID, to better match the AppStream specification.
* trivial: Rename an internal methodRichard Hughes2014-08-211-2/+2
|
* trivial: Show the full application ID in more placesRichard Hughes2014-08-191-1/+1
|
View Lorry Controller Status