Debian Patches
Status for qt6-declarative/6.11.2+dfsg-4
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| upstream_71381b2_fix_js_crash_sparc64.diff | V4: Sign-extend heap pointers on 64-bit SPARC Linux maps user memory above the VA hole on sparc64, from 0xfff8000100000000 on T4 and M7 and from 0xfff0000100000000 on M8 and UltraSPARC III, so the heap pointers have their top bits set. The sparc64 encoding kept bits 0-51 and required the rest to be clear, so the engine crashed during start-up. Store the low 57 bits as the 5-level page table layout does, and sign-extend from bit 56 when decoding. That covers the 44-, 48-, 52- and 53-bit VA layouts the kernel uses on sparc64. With GCC this branch is only compiled once qtbase defines Q_PROCESSOR_SPARC_64 for it, see the qtbase commit "qprocessordetection.h: Detect 64-bit SPARC when building with GCC". Amends commit c7722d4ed61d6a887e9f6c403ffa10b2048de2a4. |
Stian Halseth <stian@itx.no> | no | 2026-10-08 | ||
| upstream_7214a32_plasmashell_hang_shutdown.diff | Fix teardown hang with QuickPixmapReader QQuickPixmapReader is a QThread subclass that encapsulates the thread usage in the constructor/destructor. The current code is broken if the QQmlEngine (and thus QuickPixmapReader) are deleted as children of the QCoreApplication. At this point no new events are posted and the deleteLater doesn't reach the relevant thread. We then hang forever waiting on a thread that will never quit. The eventloop_hack was added in 2010 (ff983ee8257b79143911ca58ebedc49b1660a0c4 before Qt split) with the comment that QThread::quit was not thread safe. Current docs say it is. |
David Edmundson <davidedmundson@kde.org> | no | 2026-10-01 | ||
| upstream_2efb7c6_correctly_compare_composites_with_multiple_engines.diff | QML engine: Correctly compare composites when multiple engines are used Since 6.11, we allow multiple types to be active for the same URL. While we converted multiple places to handle this, we missed adjusting the property validator. There, we just picked the last inserted property cache, which can lead to validation failures. Fix this by adding a fallback path which checks all potential matches when the initial lookup would yield an error. The same kind of check is also needed in qmlobject_can_qml_cast. Amends 9cc23ca91c404f24fff3c36e8c9425ae674f341d. Change compared to 6.12: Dropped the isComposite check, which doesn't exist there. That causes one potential extra unneeded lock acquisition, but only in a case where'd we get an error anyway. (cherry picked from commit 94a76ed58629932f6d689c48ecf8d3132188923b) (cherry picked from commit 8c23f65ec4889f9ff7f609002a129698796de2b8) |
Fabian Kosmale <fabian.kosmale@qt.io> | no | 2026-08-31 |
All known versions for source package 'qt6-declarative'
- 6.11.2+dfsg-4 (sid, forky)
- 6.8.2+dfsg-7 (trixie)
- 6.4.2+dfsg-1 (bookworm)
