When QAbstractItemModel::data() returns an invalid QVariant (representing 'undefined' in JS) for a role bound to a typed property like 'required property string foo', the property-to-property binding's write was silently failing. This happened because writeValueProperty() only uses QMetaType conversions which cannot convert an invalid QVariant to a QString. The property kept its old value, and subsequent valid writes of the same value would not trigger change signals. Update QQmlPropertyToUnbindablePropertyBinding::update() to perform proper type coercion using metaTypeToJS() and metaTypeFromJS(). Pick-to: 6.11 Fixes: QTBUG-145016 Change-Id: I8421e813ddde50d4c252907045448b22a36a0a70 Reviewed-by: Olivier De Cannière <olivier.decanniere@qt.io>
When a QQmlIncubator is destroyed during statusChanged(Loading), the QRecursionWatcher doesn't detect this (it only detects nested changeStatus calls). This caused incubate() to still run, creating an object (result) that no one would ever free. Check p->q before calling incubate() to avoid that. When required properties are unset, the error path in incubate() creates the object (result) but never deletes it when errors occur. Delete result in both the Execute phase error path and the finishIncubate error path. This requires a small fix to QQuickStackElement. QQuickStackElement would repeatedly try to initialize an item when the initialization had already failed and then inconsistently early return on the second attempt, leaving a dangling pointer around. Check the "init" flag rather than "item" to determine whether we should initialize. Pick-to: 6.11 6.10 6.8 Change-Id: I7bc47e4e02011d4a8e04f8b5da702bfc62f0c0a3 Reviewed-by: Sami Shalayel <sami.shalayel@qt.io> Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
Roles act like properties. When a role has the same name as a property it constitutes an override. Currently we preserve the old behavior and simply allow such overrides, making an override of a property by a role as final, we might want to warn about them in the future. amends a910cd9f1379e87953922a7642f313ef980ae9a2 Change-Id: I91be241109c8eb017c5b657292406dd831afd623 Task-number: QTBUG-98320 Pick-to: 6.11 6.11.0 Reviewed-by: Mitch Curtis <mitch.curtis@qt.io> Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
There are too many of those in Qt code and we haven't properly announced our intention to warn about shadowing. Task-number: QTBUG-98320 Pick-to: 6.11 6.11.0 Change-Id: I048b555131c2912972362626f3f39077cc9df656 Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
This commit enables expanded Override Semantics for the properties added from the MetaObject path. It also introduces logging categories, so some of them could be silenced when they are not actionable for the user Task-number: QTBUG-98320 Change-Id: Ifb414ab795bc78f23a151666bb0d8af9b7a18d76 Pick-to: 6.11 Reviewed-by: Sami Shalayel <sami.shalayel@qt.io> Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
The return value from releaseObjectWeak() is generally used to determine whether we can get rid of the cache item. Before splitting the object reference it checked the composite of both kinds of references. Restore that behavior by checking both explicitly now. Amends commit e4db77906b3f8bdcbe2ee20ff62ad818c6817e24 Pick-to: 6.11 Fixes: QTBUG-143733 Change-Id: I009ed93ea0d56b19f530175db6e9156deb7afaab Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
emitChanges() may release cache items. If we still hold a strong object reference at that point, the reference counts can get out of sync. Amends commit e4db77906b3f8bdcbe2ee20ff62ad818c6817e24. Pick-to: 6.11 Fixes: QTBUG-143262 Change-Id: I17cea127528685ecd1052f61902c2880cc9f1f84 Reviewed-by: Sami Shalayel <sami.shalayel@qt.io> Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
Since QVector is just an alias, let's finally do what we could have done for Qt 6.0. Pick-to: 6.11 Change-Id: Ia9dbc8ecf4f099339cf7639acb22c14b24fa680b Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
The callback can do pretty much anything, including another recursive call to drain(). We need to make sure we're done with the container and its iterators before we let that happen. In the common case of maxPoolTime == 0, this is really easy. Simply swap the container into a temporary and drain that. The case of maxPoolTime != 0 is a bit more complicated, but the algorithms library gives us a way to do it with std::stable_partition and std::move that should be good enough. Fixes: QTBUG-102811 Pick-to: 6.10 6.8 6.5 Change-Id: Id67a9b9c2a7279536e5c2a315b139fcd8024e243 Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io> Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
When writing through a delegate, the original data is already written. We are only missing change signals. Pick-to: 6.10 Task-number: QTBUG-139941 Change-Id: I3124a8dc5852e3858a1e9bd9b74c07608bb5e446 Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>