[qt/qt/qtbase]: Summary of bulk changes made

Git repository change summary for qt/qt/qtbase
Pushed by mirror-service into branch 'dev'.
Changed from 823afbe2fee824e048939d278e712b6b78fc9999 to c5fc610a093e7e6b7ce84007f475271d999b265a
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 10f060ec50a17fc85fe4e7f6ecbf26a7e501962a by Thiago Macieira (on behalf of Alexey Rochev) on 19/08/2026 at 00:55..
QNetworkProxy: support socks5h scheme with libproxy backend

The user might set *_PROXY environment variables to a proxy URL with socks5h scheme, we need to support it.

[ChangeLog][QtNetwork][QNetworkProxy] The class now supports a scheme of
"socks5h" in *_PROXY environment variables, when using the libproxy
backend on Unix systems ("socks5h" indicates the SOCKS5 server is
capable of resolving hostnames).

Pick-to: 6.12
Change-Id: I0939b969c01cbb408c92224f9537d1187e5e8c86
Reviewed-by: Thiago Macieira <[email protected]>
Reviewed-by: Mårten Nordheim <[email protected]>
https://invent.kde.org/qt/qt/qtbase/-/commit/10f060ec50a17fc85fe4e7f6ecbf26a7e501962a

Git commit 23914c08819684aa44e85c74eaa23abd49f7f2d1 by Alexey Rochev on 19/08/2026 at 00:55..
QNetworkProxy: don't break proxy capabilities for socks5h proxy with generic backend

We don't need to add HostNameLookupCapability since it's the default for Socks5Proxy,
and doing so by calling setCapabilities(QNetworkProxy::HostNameLookupCapability) erases all other capabilities.
Just treat socks5 and socks5h schemes the same.

Change-Id: I9c7037fd11ae828376bd87014b528b4e421022c5
Reviewed-by: Thiago Macieira <[email protected]>
https://invent.kde.org/qt/qt/qtbase/-/commit/23914c08819684aa44e85c74eaa23abd49f7f2d1

Git commit a22ce9f4123c8fe4381b7889de8b415e2450e94d by Marc Mutz on 19/08/2026 at 02:24..
QArrayData/Ops: defend against -ftrivial-auto-var-init

Clang 21 confirms¹ that it inserts memsets/stores due to
-ftrivial-auto-var-init in these four places.

The pointer variable is written by each allocate*() function (they all
call allocateHelper(), which, as the first statement, assigns nullptr
to them. We could just let it be, except that in the Clang reporting,
this shows up in _every_ TU. 16k times in my builds, which is far from
a complete qt5 one.

The others are ... more interesting.

It turns out that -ftrivial-auto-var-init does _not_ just handle
_trivial_, but _all_ automatic variables. Maybe it's one of the other
flags that we pass for hardening, but Clang itself attributes it to
-ftrivial-auto-var-init.

It seems to only omit initialization when it can prove the constructor
of the automatic object writes to these bytes. If the constructor body
is not visible, e.g. because it is out-of-line, then Clang memsets.

These four sites were flagged by Clang itself, so while there could
be more, these are definitely affected.

QList etc are not in the business of second-guessing the
initialization behavior of unrelated types, so prevent the memsets by
using __attribute__((uninitialized)). Note that this affects only the
stacked temps. The final value that ends up in the container is using
placement new, is not an automatic variable, and therefore already not
subject to -ftrivial-auto-var-init.

The copy ctor is obviously not our concern. But the two `tmp`s are
always at least value-initializing, even if the args... are empty,
never mere default-initialization, so we don't change anything for
trivially constructible types, either.

The one thing we lose is padding initialization. But we want other
tools to tell us, not "hardening" optiona papering it over.

Amends 9ff1e6d80bbd5b44b9ec4c0a837d9a4c962698e4 (6.8).

¹ -fsave-optimization-record=yaml \
  -foptimization-record-passes=annotation-remarks

Pick-to: 6.12 6.11 6.8
Change-Id: I342172cf69299b8d9cd73c39b98951edc3149339
Reviewed-by: Allan Sandfeld Jensen <[email protected]>
https://invent.kde.org/qt/qt/qtbase/-/commit/a22ce9f4123c8fe4381b7889de8b415e2450e94d

Git commit 4563e95ae94898ac3936ca8bb636aadde19d3583 by Marc Mutz on 19/08/2026 at 02:24..
QHttpHeaaderRange: fix EQUALITY_COMPARABLE macro

This is a literal type, so use the corresponding macro to make the
relational operators constexpr, too.

Found in API-review.

Pick-to: 6.12
Change-Id: I815963f3e88af86ad57ededec5ee359d43ba0dc0
Reviewed-by: Mårten Nordheim <[email protected]>
https://invent.kde.org/qt/qt/qtbase/-/commit/4563e95ae94898ac3936ca8bb636aadde19d3583

Git commit d45c12417902975a68a265e0343dc65d4ba3dd0e by Edward Welbourne (on behalf of Marc Mutz) on 19/08/2026 at 02:26..
QtTemporalPattern: simplify supports()

We don't need to distinguish between date and time categories. We can
use a single bit-field for both. This wastes only 94-40-24 = 30 bits
of space, so four bytes, and gets rid of the magic numbers (shifts and
sizes) in the function. It also means we can reuse the CHECK macro for
both time and date components.

As a drive-by, factor the mapping into a small helper container. It
may eventually grow to be useful elsewhere, complementing my
unfinished EnumIndexedArray.

Amends 1a9f52ad2d0d42d568ee6a3a9607f8fb878875a3 (6.12).

Pick-to: 6.12
Change-Id: Ifc6d6edbcf91afe9e0b1238dbfaf1e2afabde717
Reviewed-by: Ivan Solovev <[email protected]>
Reviewed-by: Edward Welbourne <[email protected]>
https://invent.kde.org/qt/qt/qtbase/-/commit/d45c12417902975a68a265e0343dc65d4ba3dd0e

Git commit 95e387902081b005feb095ad2dae05add556840b by Lorn Potter on 19/08/2026 at 05:37..
wasm: point to wasmdeployqt when a preload manifest is missing

A shared (dynamic) build generates an app HTML that preloads
qt_plugins.json and qt_qml_imports.json, but those manifests only
exist after wasmdeployqt has been run. When the app is served without
that step, the loader failed with a bare "Could not fetch preload
file" message that gave no hint at the cause.

Extend the error to tell the developer to run wasmdeployqt, turning a
confusing startup failure into an actionable one.

Change-Id: I8756a1ee4edd4e391b7f916ceb1dc9050d149a9f
Reviewed-by: Morten Johan Sørvig <[email protected]>
https://invent.kde.org/qt/qt/qtbase/-/commit/95e387902081b005feb095ad2dae05add556840b

Git commit c5fc610a093e7e6b7ce84007f475271d999b265a by Dennis Kim on 19/08/2026 at 05:37..
ohos: don't report a hide()-issued minimize as WindowMinimized

When hideAbility() is unavailable, QOhosView::hideMainWindow() takes
the window off-screen with a native minimize(). OHOS then emits a
MINIMIZE status, which handleWindowStatusChange() promoted to
Qt::WindowMinimized.

hide() and minimize() are orthogonal in Qt: a hidden window must stay
in WindowNoState. Promoting the status drives it NoState->Minimized,
so the next show() takes the restore branch and delivers a spurious
extra QShowEvent. tst_QWidget::eventsAndAttributesOnDestroy catches
this: show/hide/show yields 3 show events instead of 2.

Suppress it: hideMainWindow() records that it owes one MINIMIZE echo,
and handleWindowStatusChange() drops that echo instead of promoting
it. Counting the owed echoes, rather than using a flag, keeps this
correct when the asynchronous echo only arrives after a subsequent
show(), and when hide/show is repeated.

Fixes: QTBUG-148259
Pick-to: 6.12
Change-Id: Ice53706c33794b2b3c47033e13b2fe5e3ffff526
Reviewed-by: SanthoshKumar Selvaraj <[email protected]>
Reviewed-by: Seokha Ko <[email protected]>
https://invent.kde.org/qt/qt/qtbase/-/commit/c5fc610a093e7e6b7ce84007f475271d999b265a