override on destructors are not very useful anyhow, but also msvc2010
doesn't like them at all.
Change-Id: I2d5f4b0d6148c5e66d2cba604d305707269d12e0
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Eike Ziller <eike.ziller@digia.com>
Moved Document::Language, Import::Type and StaticAnalysis::Severity
enums to qmljsconstants.h and renamed values removing the redundant
part.
Thus the effective length changed little or improved
(Document::QmlLanguage => Language::Qml).
The separate file allows better reuse of enum values without introducing
circular dependencies.
Change-Id: I5186d7c04f5d3f6c289068b919be5ff1ff118326
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@digia.com>
Qmljs now keeps a cache of parsed qrc files, and can resolve "qrc:" links.
This breaks the assumption that the name that the qml files has on
the filesystem is the one that qml sees, and that contents of
directories can be found just looking at file whose path starts with the
directory path.
Currently the first file is used when multiple qrc files contain the same
path, but support for strict and weak path resolving is already there.
At the moment only qrc files for projects that call updateQmljsCodeModel
are updated.
ChangeLog: QmljsSupport: Imports using qrc links are resolved.
Task-number: QTCREATORBUG-8953
Change-Id: I695fac2692af2417d49c192c580a1c2e7b4873f4
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@digia.com>
We had a qmltypes file that was encoded in UTF16 and did not work.
We should at least warn about it.
Change-Id: I42555782ee16ddd25552f919845aa85ff1f3f636
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@digia.com>
Ran script to remove inludes on a trial-and-error basis and
manually corrected it (Parser excluded).
Change-Id: I5ec6e1076430009bb72094411b2c3386f8bea548
Reviewed-by: Aurindam Jana <aurindam.jana@digia.com>
Reduce the padding in some often-used structs for 64bit machines.
32bit machines should also profit from most changes, but to a
lesser degree.
Change-Id: Ic4c67b94e962731de4f31164c52a372d78944ccc
Reviewed-by: Tobias Hunger <tobias.hunger@digia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Getting the #include directives ready for Qt5. This includes the
new-project wizards.
Change-Id: Ia9261f1e8faec06b9285b694d2b7e9a095978d2b
Reviewed-by: Eike Ziller <eike.ziller@nokia.com>
* Don't build all default values (including the global object)
separately for each ValueOwner instance.
* Instead, keep all global, immutable values in a single, shared
instance.
While refactoring, some cases where we *modified* the global object had
to be removed:
* C++ context properties no longer get injected into the global object,
instead they now have their own scope just above the global one.
* The Qt object's prototype no longer gets modified in Link. Instead,
it's now a reference to the "Qt" object provided in a qmltypes file.
* The whole concept of a function 'Activation' that could potentially
affect the global object was removed.
Change-Id: Id382faf965efa747fcc7a9b0bc2c90429d84d61b
Reviewed-by: Leandro Melo <leandro.melo@nokia.com>
* FunctionValues know about optional arguments (for builtins)
* ASTFunctionValues only report themselves as variadic if they
use the 'arguments' array.
* Function argument hint shows optional args and variadic.
* Completion automatically adds parentheses.
Change-Id: Ib2598600ff8b1ce8c5de3bcabd24a3e171ff3a57
Reviewed-by: Roberto Raggi <roberto.raggi@nokia.com>
The qmltypes files shipped with libraries shouldn't, but could define
components that form a prototype cycle. Use of the new
QmlObjectValue::prototypes() function avoids the code model getting
stuck in an infinite loop for these cases.
For plain ObjectValues this is already solved by using
PrototypeIterator.
Change-Id: I0f562672ab99b454bc3a0121c7a1b9eca43476bb
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@nokia.com>
To distinguish known-to-be-undefined from a genuinely unknown value.
Change-Id: I606b4ea4d726f94553400b8950d3c0a4e76564a8
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@nokia.com>
And introduce a good error message for the case when someone
forgets the specialization of the value_cast template.
Change-Id: Iec55a839e8f5eef5872b1dab8601f66e0e0c88de
Reviewed-by: Leandro T. C. Melo <leandro.melo@nokia.com>
This means the code model will now offer correct completion and
highlighting for arguments of signals in their handlers, example:
MouseArea {
onClicked: {
mou<complete> // now also completes 'mouse'
}
}
Reviewed-by: Fawzi Mohamed
Change-Id: I01838ef00e391b13e6e5a832c9ec3cd983689c5b
Reviewed-on: http://codereview.qt-project.org/6147
Reviewed-by: Christian Kamm <christian.d.kamm@nokia.com>
Sanity-Review: Christian Kamm <christian.d.kamm@nokia.com>
They need not to be identical. Additionally:
* Rename ImportInfo::name to path, because that's what it does.
* Add ImportInfo::name, for getting the uri with the names
separated by dots.
* Allow for exportMetaObjectRevisions in qmltypes files.
* Allow for exports with an empty type name, as generated by
qmlRegisterRevision. They are used for associating meta object
revisions with non-exported types.
* Rewrite the Qt 4.7 import to QtQuick 1.0 at an early stage. In
preparation for the Qt 5 type information update, where Qt 4.7
is gone.
Change-Id: Ia287193623d9530a56b9eb8d2481d50aabd94c3e
Reviewed-on: http://codereview.qt-project.org/5309
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@nokia.com>
The problem was that if you exported
A 1.0, A 1.1 and B 1.0 where A is the prototype of B
the code model had not enough information to know that, depending
on the import, B 1.0's prototype should be A 1.1 or A 1.0.
To solve this problem QmlObjectValues now store the import's version
as well as the local component version. In the example above B 1.0
would have import version 1.1 if the 1.1 module was imported and thus
be able to choose the right prototype.
Change-Id: I7ef33f12ca5a528c62b2a8240f4b5720b0ebd4c3
Reviewed-on: http://codereview.qt-project.org/5129
Reviewed-by: Thomas Hartmann <Thomas.Hartmann@nokia.com>
* Change from Reference to FunctionValue: There was no reason for them
being References as the argument types never need lookup.
* Enumerate in global completion: When you define 'signal foo(int a)'
you now get completion for 'foo' and will get the function argument
hint popup.
Change-Id: Ic348db477a34ba468dfdb690499a9cd8fd605cd2
Reviewed-on: http://codereview.qt-project.org/4774
Reviewed-by: Thomas Hartmann <Thomas.Hartmann@nokia.com>
Since several reference lookups involve Evaluate which may cause
further reference lookups, we need to be able to pass the existing
ReferenceContext to avoid cycles.
Change-Id: I2f1eeaad4d6b6ff094413d51077b03c985f6fab4
Reviewed-on: http://codereview.qt-project.org/4653
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@nokia.com>
* Also look for enums in prototypes.
* Report back which QmlObjectValue had the enum.
* Fix a bug where enum lookup was always skipped.
Change-Id: I9c9fd8f8cf9bd8cc5f1bb5688fef5786267cd794
Reviewed-on: http://codereview.qt.nokia.com/4192
Reviewed-by: Thomas Hartmann <Thomas.Hartmann@nokia.com>
The distinction between QmlJS and QmlJS::Interpreter has always been
weak and the extra namespace just added an unnecessary complication.
Change-Id: I4db8ef4bd91b5f6bf610a9d23fdbf55bd60250fc
Reviewed-on: http://codereview.qt.nokia.com/2743
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@nokia.com>
Previously Context was not entirely thread safe and had to be
copied locally. Now it is thread safe and its lifetime
managed by QSharedPointer.
The non-safe parts were moved into ScopeChain in a previous commit.
Change-Id: I851a93de85cbd6391dbea0fe33b011e2e093addb
Reviewed-on: http://codereview.qt.nokia.com/1695
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@nokia.com>
Context is created by Link and has information about imports
for all Documents in a Snapshot.
ScopeChain represents how lookup is done at a specific place in
a Document.
Change-Id: I874102d57bbaf1a497fa3f27633bed6ee75dcf10
Reviewed-on: http://codereview.qt.nokia.com/1694
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@nokia.com>
