Use the correct type to look up the sub property.
Task-number: QTBUG-24606
Change-Id: Iab372ca7440249241bf855d2875a947ad8fe4527
Reviewed-by: Chris Adams <christopher.adams@nokia.com>
Reviewed-by: Roberto Raggi <roberto.raggi@nokia.com>
The test is blocking the CI because it crashes on linux/x64.
By "luck" it doesn't crash on the other test machines -- valgrind
still complains, though.
Task-number: QTBUG-24734
Change-Id: I6bc12ca72f8900339bc9139b40cf828aff8e47ba
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
The worker_remove_element test calls processEvents() before
calling waitForWorker(). It's possible that the worker actually
finishes during the processEvents() call. In such a situation,
waitForWorker() should return right away; otherwise it would
wait for 10000ms for a signal that had already emitted, and the
test would fail.
Change-Id: I8e98a3297cf5f360c1c405b1baa7524cc6593d81
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
QList<Type>::reserve() is used upfront to allocate necessary memory in a one
go. This tells us straight away whether allocation is possible at all and
reduces re-allocations and consequent memory copies.
This also has the side effect that no spare memory is allocated, also allowing
up to (and including) INT_MAX elements to actually be stored in the underlying
QList, as long as enough memory is available to satisfy the allocation request
and subsequent fill.
The qqmlecmascript::sequenceConversionIndexes was changed to not attempt
INT_MAX allocations as, given enough memory and virtual address space, that
might succeed but take a really long time.
Change-Id: I4b0c965e9c23be78874343a70d7c155933c80903
Reviewed-by: Chris Adams <christopher.adams@nokia.com>
Reviewed-by: Kent Hansen <kent.hansen@nokia.com>
If a listmodel with static role types is created, it would crash
if a role was assigned a value type such as string, and then
subsequently assigned an array (sub list) value from a dynamic
meta object (created when using get() from JS).
Change-Id: Ibfd0b0b40be13b04103b49462cfae42a5c9f9fb9
Reviewed-by: Martin Jones <martin.jones@nokia.com>
Remove QQmlDebugClient and relevant classes from the
library and move to client code.
Change-Id: I6f526b3f0c92970dcad5e5abd8585bb9b406349e
Reviewed-by: Kai Koehne <kai.koehne@nokia.com>
Since we've been changing from QtDeclarativeDebugServer to
QQmlDebugServer anyway, we might as well change it to the more
readable "QML Debugger:" prefix.
Change-Id: I852577233d7cdb1f57adc43ec1b85a14d212574d
Reviewed-by: Aurindam Jana <aurindam.jana@nokia.com>
Revert the names of the services changed in
b855240b78 to
maintain BC.
Change-Id: I79826d92fd09c41e0020541b0c7bb77b5b2ecb1f
Reviewed-by: Kai Koehne <kai.koehne@nokia.com>
The type "real" was documented to be a single-precision float, but
that's incorrect. It's always been double.
However, signal parameters of type "real" would be mapped to the C++
type "qreal", which can be either float or double depending on the
platform.
Since JavaScript floating point numbers have double precision, QML
should use the same, to avoid potential loss of precision.
With this change, "real" behaves the same as the QML "double" type
(which already guaranteed double precision). Even though it's
redundant, "double" is kept to preserve compatibility.
Added tests for the "double" type to the QML meta-object autotest,
and a test for the "real" type that ensures there's no loss of
precision.
Change-Id: I1a77f1fd45082ff670684a935e17d1a46ea75d84
Reviewed-by: Aaron Kennedy <aaron.kennedy@nokia.com>
- Changed tests to use TESTDATA
- moved qqmlcontext to private test as it contains private header
- added check for cross_compile option to skip when sources not available
Change-Id: I0f68f58ffcb1b41b8e40a9851e3e003fe72ee2f9
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Reviewed-by: Michael Brasser <michael.brasser@nokia.com>
Move the code dealing with QImage and QPixmap out of QQmlEngine
and into the QtQuick library. QQmlEngine remains the owner
of image provider resources, but does not use them directly.
Change-Id: I52083581394d9c308db446372883eb7479ccf807
Reviewed-by: Martin Jones <martin.jones@nokia.com>
This patch maintains the old function-assignment behaviour for non-var
properties, and will be reverted soon.
Change-Id: I523e464501106616c51ff7478f7eb7171c1ca350
Reviewed-by: Matthew Vogt <matthew.vogt@nokia.com>
This commit changes the semantics of function assignment in QML.
Previously, function assignment was interpreted as binding assignment.
Now, function assignment is interpreted as function assignment, and
therefore fails for all property types other than "var" properties.
To support imperative binding assignment, a new function was added to
the Qt object: Qt.binding(function) which takes a single function
parameter and returns a function object which will be interpreted as
an assignable binding expression by the QML engine.
Finally, this commit also slightly changes the semantics of var
properties in that the "special" JavaScript values of null and
undefined may be assigned to var properties, rather than being
interpreted as reset requests.
Task-number: QTBUG-21842
Change-Id: Iee99a878b9badf0fb76e983da7ebfa493f55ceb5
Reviewed-by: Matthew Vogt <matthew.vogt@nokia.com>
Symbols beginning with QDeclarative are already exported
by the quick1 module.
Users can apply the bin/rename-qtdeclarative-symbols.sh
script to modify client code using the previous names of the
renamed symbols.
Task-number: QTBUG-23737
Change-Id: Ifaa482663767634931e8711a8e9bf6e404859e66
Reviewed-by: Martin Jones <martin.jones@nokia.com>
Chris Adams
Michael Brasser
Lars Knoll
Kent Hansen
Matthew Vogt
João Abecasis
Glenn Watson
Aurindam Jana
Kai Koehne
Kurt Korbatits