This collides with injected signal parameters. qmlcachegen cannot tell
those cases apart.
[ChangeLog][QML][Important Behavior Changes] QML warns about JavaScript
variables being used before their declaration now. This is almost always
a mistake. It is particularly dangerous in the presence of injected
signal parameters because qmlcachegen cannot identify a name collision
between an injected signal parameter and a variable being used before
its declaration. It therefore miscompiles such code. You can turn off
the deprecation warning using the "qt.qml.compiler" logging category.
Pick-to: 6.1
Task-number: QTBUG-89943
Change-Id: I8a9424ca8c6edd562402fe5c560ba7e8344b5585
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
The "2" is meaningless and there is a better name available now.
Task-number: QTBUG-85064
Change-Id: I65d26b06712ed7dcf2825f16dffaa6060dd86985
Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
We cannot use reinterpret_cast here, because testUnit->constants()
returns a pointer to quint64_le data, which needs to be converted to
native endianness first.
Here I used the QV4::Value converter that accepts quint64, so this
conversion now happens implicitly.
Change-Id: Iff6e3e4554af8890e61cb06e6fd79339c7a14653
Pick-to: 6.0 6.1
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Currently removing an item connected to a particle
only invalidate the particle that will be reused for
the next item set. This has the effect that the new
item inserted appears in the same position as the
old one just removed. This patch force killing the
particle to assign next item to a new particle
starting from initial position.
Pick-to: 5.15 6.0 6.1
Change-Id: Ic937a6d7aea65368cceb1405bb81ef1502d988a5
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Previously this just failed silently and returned garbage data.
Change-Id: Ia8b72836aa0ccfd50fd18b0f813c2bf3a00801c5
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This way it's fundamentally impossible to add the same animation job to
two different group jobs. The pointers are not exposed anymore and no
one can re-order the jobs.
Task-number: QTBUG-90401
Change-Id: Iebff4b64960c853915dd32714acd144fc5cdc00d
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
setCurrentAnimation() can indirectly delete the animation group job
itself by invoking the animation controller. Use the RETURN_IF_DELETED
mechanism to avoid the resulting dangling pointers.
Task-number: QTBUG-90401
Pick-to: 6.1 6.0 5.15 5.12
Change-Id: Ibd0ad21e8d3af4760604c3ff37dc46101d5f49ad
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
More tests fail on offscreen platform when tested on QEMU on CI.
Pick-to: 6.1
Task-number: QTBUG-63185
Change-Id: I293e9b32078bf2567fbb9773cedf6777ad182a69
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
This is separate from the fix, since the test is supposed to also be
merged into dev. (Where the fix was not needed)
Task-number: QTBUG-86567
Pick-to: 5.15
Change-Id: I2cf1a4b11eed4fe356588aeff322d3a432f0fe83
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This patch ensures that the QML Binding element can also save and restore
C++ bindings. Should QQuickItem's x and y property be ported to the new
property system, we'd need new test cases to verify that "old-style"
bindings are still handled correctly. This task is however left for the
change porting the properties.
Task-number: QTBUG-90493
Change-Id: I506ffa1060ff32a7d722214e5ccd469bdaa61ff8
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
If a QML component wants to mark properties of its "parent" component as
required, it can do so via
required propertyName
The information about those properties is stored in a
RequiredPropertyExtraData data structure. This structure is already
serialized to disk in the QQmlIRWriter. However, we neglected to restore
it so far in the loader.
Fixes: QTBUG-90538
Pick-to: 5.15 6.0
Change-Id: I789daff9bc881e4f35c942c77f5116b5284de81b
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Default properties are always local. There is no way to declare a
default property for a foreign type as the default property is queried
directly from the classinfo at runtime.
Change-Id: I30efb6fba190957ac2a4ad86da437f209cd1f3ad
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
No one should mess with those. All the pointers have complicated
ownership semantics. We can just befriend the test instead of making it
all public.
Task-number: QTBUG-90401
Change-Id: I6c4adbab7046b40db7f4628780ef928445ea3eb2
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
If we generate a local anonymous type, then that's the local part of a
QML_FOREIGN local/foreign couple. Any QML.Extended in there belong to
the foreign type.
Change-Id: Ic1706045eff03dd7b1b553240596ffc21818c8bd
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
In TestCaseCollector::enumerateTestCases, we visit the super compilation
unit of QML tpyes to check if they might be instances of TestCase.
However, in the case of inline components, the super unit is the current
compilation unit, and we would recurse endlessly.
This does not address the issue that an inline component might actually
inherit TestCase. However, as this only affects the enumeration output
and does not actually affect test execution, this is not that much of an
issue. It should also be noted that the enumeration also fails in any
case where TestCases are loaded dynamically (with a loader), so the
method is not 100% accurate even in the absence of inline components.
Fixes: QTBUG-90740
Task-number: QTBUG-90762
Pick-to: 5.15 6.0
Change-Id: I7e133d62c4f62fc46e9bd3999ff755f7ded3c386
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Previously, the assumption was that each object could only have a single
extension object. As proven by the new qqmllanguage test this is not the
case. Each registered object in the type hierarchy can have its own
extension. Therefore, adjust the algorithms that generate qmltypes and
iterate the extension objects when analyzing them.
This leads us to the realization that anonymous types can in fact
meaningfully carry extensions and implement interfaces. Adapt
qmltyperegistrar accordingly.
For the test to compile, however, we need to realize that the class
declaring interfaces needs to befriend all potential subclass's
QmlInterface structs. Fix that, too. The rabbit hole went deep.
Change-Id: Ia451897e927e03b95c3062e829edf1dfcd216613
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
It's apt to be more accurate and may even be optimised.
Comment on a benchmark where we could use Math.hypot(), but that would
break comparison with Qt 5 results.
Change-Id: I7c37dd3df82fdef18e7ebb0e1548198afd256faa
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This simply exports the name which are already available in the json
files generated by moc. We do not consider whether the methods are
non-private for now. MEMBER is not supported either, but might be added
if the need actually arises.
Fixes: QTBUG-90711
Change-Id: If3ee18c8ce60499676a7ee22df569cba0912e22f
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
The "in" operator may throw an exception.
Change-Id: I7d0b6e2212ac6ec237fbf14719349f8e23810028
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
The test failed reliably when the directory in which the test started had
the same number of files as the directory that the test wants to read.
That's because the QTRY_COMPARE matches immediately and doesn't process
events, which leaves queued signal emissions pending. So, count tests
passed - for the wrong reason - and follow up tests failed.
To make the test robust, start with an invalid directory, which we know
is empty. Once switching into the test directroy, the test will have to
process events.
Remove the BLACKLIST file.
Fixes: QTBUG-90468
Change-Id: I9b3c4dc1a15b0b5ab6c632c12752b038164b9d9d
Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
If we don't return there, the newly imported script is promptly
overwritten by an invalid module.
Change-Id: I788a7275e2c190a20c176da35f5c76ac9f6ad02b
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Before, we would prepend the namespace to the name, but then continue
right away, never checking the new name.
Change-Id: If90db7d33536fb4b549321c2d6b677040605b6f0
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We need to be able to resolve any QML type from its C++ name using
QMetaType::fromName(). qmltyperegistrar can generate the missing
metatypes, either by creating synthetic ones (for namespaces), or by
making sure the existing ones are registered (for others).
Change-Id: If775af56d891f2c2a5bb94589b3cb05a199c7c35
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
This change ensures that bindings created in QML between new-style
properties contain information about which property caused the loop. To
do this, we store additional information about the property involved to
retrieve its name and position at a later point. We print the warning in
case we detect a binding loop in evaluate, and also set the error
reporting callback correctly, so that the condition can be reported when
the loop is detected in another part of the binding evaluation.
In addition, we do not only set the QPropertyBinding's error member when
JS evaluation results in an error, but also print the warning with
qmlWarning.
Fixes: QTBUG-87733
Change-Id: Idb25237d1f57355ca31189e6bf2a918430b3a810
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
It makes little sense that you can construct an error object but not
throw it. The test was definitely meant to actually throw the error
object, not return it.
[ChangeLog][QtQml] QJSEngine has gained an additional overload to the
throwError() method, with the effect that calling throwError() with a
character literal as argument is now ambiguous. You should explicitly
construct a QString instead.
Change-Id: I90c6c9edf10509daa142a86581d6a3f7ff45af2c
Reviewed-by: Maximilian Goldstein <max.goldstein@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
As all QRhi stuff is private, it needs to have a public counterpart in
Qt Quick in order to fully support the case of Vulkan-based
QQuickRenderControl usage.
Change-Id: Iaf9a7aa56022acd31af6ebf16de6b83a04966ff4
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
All those types are properly defined in the qmltypes files now. We just
need to search the enumerations the same way as methods and properties
in order to find everything.
Also, deduplicate the code that resolves properties, methods, and enums
by using a common template for iterating the scopes.
Change-Id: I0bf1423974d0ec8f602ecd0342522b3e981a8586
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Old API assumes sender == receiver, which results in wrong handling of
connections when receiver is deleted: connection is not removed or
notified elsehow as it's not really tied to a valid receiver
Task-number: QTBUG-86368
Pick-to: 5.15 6.0
Change-Id: I0f3115f1b0f26cf353752ba2b8fd88e0f3bdd388
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This is a partial revert of 78ab4b8d8b, as
the fuzzing infrastructure has not been migrated to CMake yet.
Change-Id: Iebd873ec3760c5394bc248ed853b57e168d0f05c
Reviewed-by: Robert Loehning <robert.loehning@qt.io>
Previously only the foreign type was searched for a create method.
Now the wrapping type can also contain it.
Change-Id: I05fb9e0c0a54c14530eb9adcae5a44df5c208be3
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
For some reason, we do not get the correct line number on macOs.
Task-number: QTBUG-90448
Change-Id: I8dc45011c68e2cbe0d30f4a54f48a74a5a6b0271
Reviewed-by: Maximilian Goldstein <max.goldstein@qt.io>
And do not choke completely on extended.
Task-number: QTBUG-90448
Change-Id: I4ac5742ec70f5ba1ed1403be444d9cc7229830c2
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
We generate the extensions as separate types. This also covers the case
of value types being extended by "themselves". We can properly express
this now, so we don't need the hackery of generating the local members
of a type with QML_FOREIGN into the foreign type anymore.
This also fixes interfaces from local types being written for foreign
types.
Fixes: QTBUG-89501
Change-Id: Ic76acd7eef09a92c1e36bd7a649c7a2deb24597b
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We special case writing functions to properties, only allowing assigning
them to var properties and QJSValue properties.
This would however break when aliases are involved. This commit fixes
the issue by resolving the alias, and then checking and writing to the
resolved property.
Fixes: QTBUG-90373
Pick-to: 5.15 6.0
Change-Id: Ia09ebe92feeaf8359c99ff9aeadc676b9fcfaa07
Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Changes handling of hovered so that the property is still updated on
disabled items, so that other items can bind to it. This is in
particular useful for tooltips.
[ChangeLog][Behavior Changes] QQuickItem::hovered will now update even
when the item is disabled.
Fixes: QTBUG-30801
Pick-to: 6.0
Change-Id: Id17298f657d7631b0e5019138ba33a7d5f863475
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Remove all qmake project files, except for examples which are used to
test that qmake continues to work.
Change-Id: Ic4abb72dc2dcd75df7a797c56056b6b3c5fe62ac
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Ulf Hermann
Alexey Edelev
Maximilian Goldstein
Dmitry Shachnev
Fabio Falsini
Samuli Piippo
Jan Arve Sæther
Fabian Kosmale
Edward Welbourne
Volker Hilsheimer
Laszlo Agocs
Andrei Golubev
Allan Sandfeld Jensen