Task-number: QTBUG-151040 Change-Id: I899b2c662e4de5bfe3f4db5af608dfb4a4ddcc93 Reviewed-by: Semih Yavuz <semih.yavuz@qt.io>
Don't let users of DomEnvironment call updateLoadpaths() and setResourcepaths(): instead, set import and resource paths before loading a file to avoid setting them during the loading of a file. Modifying the import and resource paths on the QQmlJSImporter while it is being used to load a file leads to the crash "A" as seen in QTBUG-149089, where a thread using QQmlJSImporter does setImportPaths(), clearing internal caches in QQmlJSImporter while a second thread, using the same QQmlJSImporter, is actually loading a QML module into the same cache. Moving the updateLoadpaths() and setResourcepaths() calls outside of the file loading mechanism avoids this situation altogether, without having to add mutexes to QQmlJSImporter. The same cause very probably also causes crash "B". Pick-to: 6.12 6.11 6.8 Fixes: QTBUG-149089 Change-Id: I1aa89419d8b78c4cee08ffd7b10116125135f5f1 Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
m_nestedFunctionDepth is a parameter that tracks whether
FunctionDeclaration is a qml object member or a plain JS local function.
While QQmlDomAstCreatorBase::visit(UiScriptBinding)
increments it, visit(UiPublicMember) never did the same. This caused the
code like property var c: { function test() {} } misidentify the inner
test as a top-level qmlmethod pushing the MethodInfo node onto the node
stack that nothing ever popped, which eventually crashes due to stack
imbalance.
Increment/decrement m_nestedFunctionDepth around UiPublicMember's
statement the same way it's already done for UiScriptBinding.
Pick-to: 6.12 6.11 6.8
Fixes: QTBUG-149244
Change-Id: Ia2e7ee82bc0170770abbfd786d88c039a3a7fa21
Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Removes unused container leftovers. In tst_qmldomitem.h just commented-out l4 since it's used in other commented-out code. Task-number: QTBUG-148221 Change-Id: I849c72729a918ebe05208e67fe2fe55bb2de08db Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Make addError() set the isValid flag to false. This makes qmlformat reject invalid files, instead of trying to format them, for example. Adapt a test that was triggering addError() but that didn't expect the new file to be error-free. It seems that the copied DomItem references the environment that does not exist in the universe, so don't try to copy the file via the universe. This also silences some QCRITICAL messages during the test. Pick-to: 6.11 Task-number: QTBUG-142711 Change-Id: I89a29733b5b18cb1a70c35feb4a9fed3ae35c8a2 Reviewed-by: Semih Yavuz <semih.yavuz@qt.io>
It seems that Q_UNREACHABLE triggers UB at runtime when hit in a release build, so remove it and log an error instead. Ignore the invalid part. Pick-to: 6.11 6.10 6.8 Fixes: QTBUG-142711 Change-Id: I445e46a6964f072a939cd6666c15c8d1ff5a0966 Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We can use the resource files from the .qmlls.build.ini also when constructing the DOM, instead of iterating over the build and import folders. Fix and rename a test in tst_qmldomitem. Task-number: QTBUG-141707 Change-Id: I9dcbd4cd307aba251e5de1d8c00c97b22dd87eb4 Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
It only calls wrap(). Replace all usages, and change the first argument to use PathEls::Field(). This helps unclutter the DomItem interface. Task-number: QTBUG-142187 Change-Id: I3aeda82b95948dde7bb7b07ab573585ac11958a6 Reviewed-by: Semih Yavuz <semih.yavuz@qt.io>
Fix FileToLoad::fromMemory to not set an empty file path on non-existent files. Otherwise, the lazy-loading machanism will see that the factory has an empty path and will think that the factory is invalid and won't populate the file. Add a test that files with nonexistent file paths get populated. This case might happen while writing tests with snippets, where no on-disk file exists. Change-Id: I48ffd5f47a5f5f5111cfd257059093ce13ed5293 Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This commit enables scriptexpressions for DomCreationOptions::Default. QQmlDomAstCreator does not know how to handle certain script expressions, notably JS lists due to the weird way PatternElementList::accept0() is implemented. QQmlDomAstCreatorWithQQmlJSScope uses custom list iteration for types like PatternElementList to make QQmlDomAstCreator support them correctly. Move the custom list iteration functionality out of QQmlDomAstCreatorWithQQmlJSScope so that QQmlDomAstCreator can also benefit from it. Rename QQmlDomAstCreator to QQmlDomAstCreatorBase. The Base version does not do any custom iteration and can be used by QQmlDomAstCreatorWithQQmlJSScope (otherwise you end up doing the custom iteration twice when adding it to QQmlDomAstCreatorBase). Create a new class called QQmlDomAstCreator that inherits from QQmlDomAstCreatorBase and uses the custom list iteration on top of it. Add a test in tst_qmldomitem to make sure that a JS list (that contains parts which are parsed as PatternElementList) is correctly loaded in the DOM. Furthermore, add minimal support for WithStatements in QQmlDomAstCreator to avoid weird test failures because of writeoutchecks in tst_qmlformat_cli::settings that uses a file with a with-statement. Also fix the qmldom tool to correctly output "null" values in JSON format by simplifying DomBase::dump()'s handling of values, and add a test for it. Not outputting "null" leads to invalid JSON generation and a failure in tst_qtquickview_signallistener at build time on android in the CI. Change-Id: I7f9258921d8515eb6b3f9cf5dbb2539ffb980aa7 Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>