[qt/qt/qtapplicationmanager]: Summary of bulk changes made
KDE Git Services - Bulk Change <[email protected]>
| Newsgroups | gmane.comp.kde.cvs |
|---|---|
| Message-ID | <[email protected]> |
Git repository change summary for qt/qt/qtapplicationmanager Pushed by mirror-service into branch 'dev'. Changed from d3a3ed947848c3f550856275f7aa3df514a84e9c to a27891c91719507b560f58da3a011746f0bfcfde Acknowledgement was received that this change introduces only existing code that has been pushed to another public open source repository. This change contains the following new commits: Git commit 54375674bede168249620ffee2f2c5c715686c58 by Robert Griebl on 13/08/2026 at 12:08.. Security review: require store signatures for System UI installs Split the task origin System into SystemDeveloper and SystemUI and derive it in one place from the D-Bus bus type and the development mode. The signature verification in InstallationTask is now purely origin based instead of consulting the development mode: * SystemUI installs always need both a store and a developer signature. Before, enabling development mode also relaxed the store signature requirement for System UI initiated installs. * ApplicationDeveloper installs (via appman-controller) reject store-signed packages and require a developer signature created by the currently set developer certificate, with the package staying within the certificate's bounds. * SystemDeveloper installs (via appman-controller) can sideload store-signed packages; a developer signature is always required. The duplicated developer-certificate bounds checks are now in the checkDeveloperCertificate() helper, which also reports a missing certificate explicitly instead of failing the package-id match against an empty certificate. The signer-binding check moved out of the verification try block, so its errors are no longer re-wrapped as verification failures. The tests install dev-signed packages via the controller origin now and a new test verifies the store signature requirement for System UI installs. Change-Id: I4c18ae36e72b7fa073b0ef62058aea855a332f6e Fixes: QTBUG-149118 Pick-to: 6.12 Reviewed-by: Thomas Senyk <[email protected]> https://invent.kde.org/qt/qt/qtapplicationmanager/-/commit/54375674bede168249620ffee2f2c5c715686c58 Git commit 2734d238fbf706a05afc267470488f4d55ee7d9a by Robert Griebl on 13/08/2026 at 12:08.. Doc: improve the taskFinished/taskFailed signals docs Change-Id: I135f163906e21fd46d2fa9cb858b80a94e283202 Pick-to: 6.12 Reviewed-by: Thomas Senyk <[email protected]> https://invent.kde.org/qt/qt/qtapplicationmanager/-/commit/2734d238fbf706a05afc267470488f4d55ee7d9a Git commit a27891c91719507b560f58da3a011746f0bfcfde by Robert Griebl on 13/08/2026 at 12:09.. Remove the Error enum and the errorCode mechanism Nothing branched on an error code except the "was this canceled" check in appman-controller, and 6 of the enum's 18 values were never constructed at all. Exception, PackageCreator, PackageExtractor and AsynchronousTask now only carry an error string. The cancellation state that Error::Canceled encoded for AsynchronousTask moves into a wasCanceled() flag in the base class, mirroring PackageCreator and PackageExtractor - cancel still wins over a concurrent real error. The "only the first error counts" guard in the two archive classes now tests m_failed instead of the code, because two of the cancel throws carry no string at all. PackageManager::taskFailed keeps its (taskId, errorCode, errorString) signature for QML and D-Bus compatibility, reporting 1 for a canceled task and the old Error::System value of 10 otherwise. This keeps appman-controller working against either side of the change; the documentation only promises 1 versus > 1. Change-Id: Id5c4575b681840a0d28a527d695545349b94c812 Reviewed-by: Thomas Senyk <[email protected]> https://invent.kde.org/qt/qt/qtapplicationmanager/-/commit/a27891c91719507b560f58da3a011746f0bfcfde