Thiago Macieira [Fri, 24 Jan 2025 19:07:58 +0000 (11:07 -0800)]
[PATCH] QLibraryInfo: speed up checking if ":/qt/etc/qt.conf" resource exists
Go straight for QResource, because this is run very early in Qt's
initialization, usually as a result of some debug message, via
QLoggingRegistry::initializeRules(). This bypasses the need to create
QResourceFileEnginePrivate, QResourceFileEngine, QFileInfoPrivate, and
QFileInfo, all of which would end up in this .isValid() call.
Additionally, I'm making it query in the C locale, which will also avoid
initializing the system & default QLocales. If a resource exists in any
language, the C locale query will find it.
Task-number: QTBUG-133206
Change-Id: I434b498903d793c12d35fffd3e297bfdbdc1b6fe Reviewed-by: Edward Welbourne <edward.welbourne@qt.io> Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
(cherry picked from commit d59e640c868f3db2d661970f3d34a22013d49053) Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>
(cherry picked from commit ae2502b4ad3d1215211bf4ed44037a40f52a313d)
Thiago Macieira [Fri, 24 Jan 2025 18:43:38 +0000 (10:43 -0800)]
[PATCH] QLocale: try to survive being created during application shut down
QLocale is very often accessed during global static destructors, so
let's try and survive if the default has already been destroyed. In that
case, we shall fall back to the C locale.
I've placed the call to systemData(), which updates the system locale,
before the initialization of defaultLocalePrivate, as the initialization
of the latter depends on the former.
Task-number: QTBUG-133206
Change-Id: I48e29b45f9be4514336cfffdf5affa5631a956a3 Reviewed-by: Edward Welbourne <edward.welbourne@qt.io> Reviewed-by: Albert Astals Cid <aacid@kde.org>
(cherry picked from commit e0a1f491567f2495443babc5aa36a038260f96c6) Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>
(cherry picked from commit bcc0e6124a2ec80df535178d056324433f9ff984)
David Redondo [Wed, 15 Jan 2025 12:52:13 +0000 (13:52 +0100)]
[PATCH] QOpenGlContext: Always unset current context in doneCurrent()
Otherwise when no other context is made current until thread exit, the
QGuiGLThreadContext destructor will try to call doneCurrent() on an
already deleted context.
It is a precondition violation to call QByteArrayView::at() with
size() as argument. The code used that, though, as an implicit
end-of-string check, assuming == ' ' and == '=' would both fail for
null bytes. Besides, QByteArrays (but most certainly QByteArrayViews)
need not be null-terminated, so this could read even past size().
To fix, use higher-level API (startsWith()), consuming parsed tokens
along the way.
Gbp-Pq: Name upstream_cve-2025-5455_fix_data_assertion_error.diff
After finding the end marker `---`, the code expected more characters
beyond: typically at least a trailing newline. But QStringView::sliced()
crashes if asked for a substring that starts at or beyond the end.
Now it's restructured into a separate splitFrontMatter() function, and
we're stricter, tolerating only `---\n` or `---\r\n` as marker lines.
So the code is easier to prove correct, and we don't need to check
characters between the end of the marker and the end of the line
(to allow inadvertent whitespace, for example). If the markers are
not valid, the Markdown parser will see them as thematic breaks,
as it would have done if we were not extracting the Front Matter
beforehand.
Credit to OSS-Fuzz which found this as issue 42533775.
[ChangeLog][QtGui][Text] Fixed a heap buffer overflow in
QTextMarkdownImporter. The first marker for Front Matter
must begin at the first character of a Markdown document,
and both markers must be exactly ---\n or ---\r\n.
Pino Toscano [Sat, 22 Jun 2024 17:55:15 +0000 (19:55 +0200)]
[PATCH] IPC: add PATH_MAX-less fallback definition for MAX_PATH
Define MAX_PATH also when PATH_MAX is not defined (e.g on GNU/Hurd).
MAX_PATH is Windows constant, and it is used in this file only in a
code path for Windows; because of this, the static fallback define
should be good enough.
Change-Id: Ic1b9fee3b62505f86aa8ec89bbd20493bfe1f67c Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Gbp-Pq: Name upstream_IPC-add-PATH_MAX-less-fallback-definition-for-MAX_PA.patch
[PATCH] XCB: Remove extra xcb_randr_select_input() calls
This fixes high CPU usage in kwin.
If there is a ConfigureNotify event for the root window, QtXCB will
call xcb_randr_select_input(). The problem is that the X server may send
an output change event in response to xcb_randr_select_input().
When kwin sees that output change event, it will process the event and
it can update its override redirect windows, which can produce a few
ConfigureNotify events for the root window and make kwin get stuck in an
update loop.
Since the QXcbConnection constructor already subscribes to the RandR
events, these xcb_randr_select_input() function calls can be removed.
Also, `true` is not a valid argument for xcb_randr_select_input(). It
expects a mask of events to listen. For what it's worth, `true`
corresponds to XCB_RANDR_NOTIFY_MASK_SCREEN_CHANGE = 1, which
xrandrSelectEvents() already subscribes to.
* Non-maintainer upload.
* Backport upstream patches to fix data races in QReadWriteLock
(Closes: #1122640).
* Backport a upstream patch to fix data races in QProcessEnvironment
(Closes: #1123679).
projects / qt6-base.git / log