summaryrefslogtreecommitdiffstats
path: root/tests
Commit message (Collapse)AuthorAgeFilesLines
* Change license for test filesLucie Gérard2024-04-045-5/+5
| | | | | | | | | | | | | According to QUIP-18 [1], all test files should be LicenseRef-Qt-Commercial OR GPL-3.0-only [1]: https://contribute.qt-project.org/quips/18 Pick-to: 6.7 Task-number: QTBUG-121787 Change-Id: I5e82161c6391caa1d44e8f3baac93a95ab80bfdb Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io> Reviewed-by: Kai Köhne <kai.koehne@qt.io>
* Enqueue new text if added while the engine is busyVolker Hilsheimer2024-03-051-2/+2
| | | | | | | | | | | | | | | | Calling enqueue() should be equivalent to say() when the engine is Ready, otherwise it should enqueue the text. The old implementation only enqueued the text when the engine was in Speaking state, overriding the current utterance when the engine was already synthesizing, or paused. Adjust test to enqueue the next text chunk as soon as the engine transitions away from the Ready state. Pick-to: 6.7 6.6 Fixes: QTBUG-122884 Change-Id: I19518a92d1ae73b01dc3de1d9ae6178f5f55b3ad Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Fix creation of locale in test_default_voice testVolker Hilsheimer2024-03-041-1/+1
| | | | | | | | | | | "en-UK" would be a (non-existent) Ukrainian regional version of English. After recent CLDR updates, this now fails, as it should, so fix it to use "en-GB" to compare with Bob's expected "Oxford English" dialect. Fixes: QTBUG-122950 Pick-to: 6.7 6.6 Change-Id: I5bc87d30b1f5f3f9804206c069be2ad5a7dc5d43 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Revert "Temporarily blacklist tst_QTextToSpeech on openSuSE"Axel Spoerl2024-01-221-7/+0
| | | | | | | | | This reverts commit 887f04a8ae406799b972f1c73b37b8e687a3e539. Reason for revert: Root cause fixed with b93bcc2c9c2880461a9aab8384c61f5ddcfa30d6 Change-Id: I86ef1bcba10e1373d21822f8377a248fdd169ca4 Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Temporarily blacklist tst_QTextToSpeech on openSuSEAxel Spoerl2024-01-091-0/+7
| | | | | | | | | | | | | | COIN VM for current openSuSE targets (Leap 15.5) is provisioned without dummy audio device. That makes all tests flaky and sayMultiple() fail. Temporarily blacklist the test class on openSuSE. Cherrypick this down to 6.5, because all versions are affected. Fixes: QTBUG-120655 Pick-to: 6.7 6.6 6.5 Change-Id: I975890fb454a09eab2039b959748e4ef78908150 Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* WinRT: fix hanging synthesize when RIFF header is removedVolker Hilsheimer2023-11-061-18/+23
| | | | | | | | | | | | | | | | Amends 428cae98619584a51c01be135fecf6b347c117b3. If we drop the RIFF header bytes, then we have fewer bytes available as well, so decrease that value. Otherwise we never fetch more data. Improve the test, which should have caught this regression, by testing that synthesise cycles correctly through the states even if no default audio output is present. The test was completely skipped because of that, not detecting the regression. Fixes: QTBUG-118668 Pick-to: 6.6 Change-Id: I3b4276c13ce2c77ec718d1b7bbca6f3421636890 Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* Fix top-level qt5 -no-gui buildKai Köhne2023-10-111-5/+0
| | | | | | | | | | | Gracefully handle the case where Multimedia is not available. Also, as we now centrally check for Multimedia availability, remove the individual checks. Pick-to: 6.6 Fixes: QTBUG-117824 Change-Id: If40c7f98f1dfa48c91f504dbdf657067044860c3 Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
* Mark the whole repo with QT_NO_CONTEXTLESS_CONNECTAhmad Samir2023-08-231-5/+6
| | | | | | | | | | By adding it to the default build flags via .cmake.conf. This amends commit c641e462e2bd33972646bd20bc76f8dff4e6d01d. Task-number: QTBUG-116296 Change-Id: Ie1e5567ef88843d2a85ec6be03cd4d72183ba269 Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* API review: remove the synthesized signal and the simple slotVolker Hilsheimer2023-05-301-11/+6
| | | | | | | | | | | | | | | They were only need for QML clients, for C++ client the version taking a callable is more convenient. But QML clients cannot really use QAudioFormat (there is no QML version of the type), and operating on QByteArrays is also not something we want QML (or JavaScript code) to do. The engines still emit a signal, as that makes it easier to implement the engine. Change-Id: Ie24a41195cd5b7e27ec2b1562fb3f8e515c5adc3 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io> Reviewed-by: Jeremy Whiting <jpwhiting@kde.org>
* API review: Allow synthesize callbacks that take a QAudioBufferVolker Hilsheimer2023-05-281-0/+11
| | | | | | | | | | | | | | Formally, QAudioBuffer is the right type for carrying audio data, but it's hardly used in Qt Multimedia itself, and not very practical to use for writing the received PCM data to a file or to stream it out to a QAudioSink (which operators on a QIODevice, e.g. with a byte array). Nevertheless, allow a callback to take a QAudioBuffer instead of QAudioFormat and QByteArray, as the QAudioBuffer facilities might be useful for some use cases. Change-Id: I260a4cf6cf91f57356373f4ef9cf248927159b40 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Add PauseResume capability flagVolker Hilsheimer2023-05-251-0/+4
| | | | | | | | | | | | | | | | | Android lacks this capability as the native TextToSpeech class has no API for this. Since we now have the capabilities enum, add PauseResume and don't list that for Android, so that applications can disable the respective UI. Do that for our examples - hide the pause & resume buttons if the capability flag is not set. Document the unsupported capabilities for all engines that lacks some features Fixes: QTBUG-113805 Change-Id: Ia8139e235f4cd968519423515e31c81285a2d349 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* API review: add word string to sayingWord signal parametersVolker Hilsheimer2023-05-241-5/+16
| | | | | | | | | | | For UIs that just want to display the currently spoken word, this is easier to connect to existing APIs. Also add the ID of the utterance, as returned by enqueue(). The index and length into the overall string stay as the final parameters. Change-Id: I70bf35eeadd24540670cad2edd42126331796f4b Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
* Test clean-up: we don't need to quote the keys of a mapVolker Hilsheimer2023-05-241-9/+9
| | | | | Change-Id: Ie4c32de764686bf0cf083ab0ad4aa20da4a53203 Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
* QTextToSpeech: rename sayNext to enqueueVolker Hilsheimer2023-05-101-2/+2
| | | | | | | | The text passed into the function is not said next, it's said at the end of the currently pending texts. Change-Id: Ic0ea885b65cdd8d7a055616fc9098d3c8cd3d397 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* TextToSpeech: make engine parameters available from QMLVolker Hilsheimer2023-04-251-0/+14
| | | | | | | | | | | | | Add a map-type property that QML code can initialize to engine-specific key/value mappings. Changing the property at runtime re-initializes the engine. Implement asnychronous initialization option in the mock engine so that we can test those code paths better. Change-Id: I0f2667b9b8e2339fa2e6966a2669f6f54ff2572a Reviewed-by: Axel Spoerl <axel.spoerl@qt.io> Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* TextToSpeech: delay engine creation, make order-agnosticVolker Hilsheimer2023-04-251-1/+55
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | With QML, the engine property will always be set after the QTextToSpeech object was created with the default engine. That QTextToSpeech's default constructure (with the engine left empty) creates the default engine implicitly is wasteful if it later gets replaced. And what's worse, the engine was the only place where property values are stored, and overriding the engine would not maintain those values. This makes TextToSpeech susceptible to the order in which declared properties are set, which breaks declarative programming. Use the special engine name "none" to delay the creation of the engine in the default constructor, and override the "engine" property to intercept calls to getter and setter. Set the engine then in the override of componentComplete. And store set values for pitch, rate, and volume in the QTextToSpeech object directly so that we can initialize the engine with those values, no matter the order in which properties are set. This also allows us to maintain the values from the old engine when changing engine. We cannot do that for locale and voice, as those are engine-dependent. However, it's not possible to set a voice directly anyway without getting it from the engine first anyway, and the voice selector is only executed at the end of componentComplete. Add test to verify that all relevant properties have the right values, even when setting or changing the engine later. Change-Id: Ib162f87c1f9ceaad1fe8674c149290ac9141fccc Reviewed-by: Axel Spoerl <axel.spoerl@qt.io> Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
* Provide a VoiceSelector attached propertyVolker Hilsheimer2023-04-231-0/+62
| | | | | | | | | | | | | | | | | | | This makes it possible to select a voice declaratively, by specifying the selection criteria in a grouped property. If the VoiceSelector is specified as part of the element declaration, then the voice is selected only once the component is completed and all selection criteria set. Since this adds a bunch of QML-specific code to the module, move the existing QML binding code and module build files into a subdirectory. Move the findVoices invokable from QTextToSpeech to the new QML-specific subclass (it's not needed as a C++ API and was documented as internal in QTextToSpeech). Change-Id: I6f8907f53b513d1108f8446d57bef5975035163b Reviewed-by: Axel Spoerl <axel.spoerl@qt.io> Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* Add QTextToSpeech::findVoicesVolker Hilsheimer2023-04-172-1/+184
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | QVoice is not a type that can be constructed by client code, it can only be constructed by the engine (as the QVoice instance carries the engine specific identifier, which an application cannot know). Applications that want to select a voice that matches a certain (set of) criteria - for instance "a male English voice" - have to go through some length to first get the list of all voices (availableVoices returns only the voices for the current locale, so one first has to go through all locales, sets each of them, and combine the lists of voices for each locale), and then postprocess that list to find a voice that matches. For C++, a variadic template function returns the list of voices that matches an arbitrary set of criteria, in an arbitrary order. Each argument is matched against the corresponding property of a voice based on its type. It's possible to match voices against only language or territory as well as a fully defined QLocale object. For QML, the function takes a map from voice's property name to value. Since QML cannot construct a locale with an "Any" territory or language, a "language" property is added to the voice type so that only the language attribute of the provided locale is compared. Make this testable by providing the mock engine with support for a parameter "voices". That parameter takes a list of std::tuple, as otherwise we'd have to create a VoiceData struct type of sorts that the test can use. Extend the list of built-in voices, and improve QVoice's debug output by printing the full locale, not just the language. Change-Id: I5266e65932ea3db52fae92a6f50caa14dbe1f2f6 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* A basic QML test case for Voice and TextToSpeechVolker Hilsheimer2023-04-165-0/+80
| | | | | | | | To verify that our APIs work from QML, no function testing performed. Uses only the mock engine. Change-Id: Ia4c28418cc5e72f6c54bcbb06e48d1d0677e73e5 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Add support for enqueuing textVolker Hilsheimer2023-03-111-0/+104
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Add a sayNext() slot that doesn't stop any ongoing speech, and instead enqueues the new text. The cross-platform implementation keeps track of enqueued texts, and processes the next text when the engine's state changes to ready. Make this the default behavior for synthesize(). It makes no sense to interrupt an ongoing process, the application can just stop an ongoing process and discard the PCM data it doesn't want. The stateChanged signal does not get emitted when the engine's state changes to Ready and there are texts in the queue. To allow applications to keep track of the text that is about to be spoken, add a new aboutToSynthesize signal that gets emitted each time text is about to be passed down to the engine. This also allows applications to make last-minute changes to the voice attributes. To accurately keep track of which text within the data structure of the application is about to be spoken or finished, applications do need to keep track of the text segments passed QTextToSpeech and update their "current" iterator with the aboutToSynthesize signal. Task-number: QTBUG-102355 Change-Id: I7b8621e15ee8d520b156e1fd771e120ded731fd8 Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Support single-argument functors for the synthesize callbackVolker Hilsheimer2023-02-281-0/+7
| | | | | | | | | Amends c03afcc297bf250baff8d0693e4db0c8cc77eeed. We already supported member functions that only take a QAudioFormat, so make this symmetrical and support it for lambdas and free functions as well. Change-Id: Ia8955ecd6ccc569aec326e469f0dd68306927218 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Allow tts engines to override capabilitiesVolker Hilsheimer2023-02-211-6/+0
| | | | | | | | | | | | Some engines might support capabilities depending on the runtime environment, such as the exact operating system version. Allow engines to override a virtual capabilities() method for that. By default, the meta-data from the plugin is used. Remove hard-coded special cases for older Android version from the test. Change-Id: I44713b3c5323f6a83713f1e2465920ab21788bab Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Add QTextToSpeech::synthesize to produce PCM data rather than audioVolker Hilsheimer2023-02-191-0/+184
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The function starts the synthesis as an asynchronous process, and emits a signal 'synthesized()' (or calls a functor) with a chunk of PCM data as a QByteArray, and the QAudioFormat in which the data is encoded. This requires a dependency to Qt Multimedia for Qt Speech for all platforms; it has so far been required only with flite and winrt backends. Implemented for all engines, except speechd and macos engines where it's not possible - these engines don't provide access to the data. The test case verifies that the implementation is asynchronous, and that it produces a reasonable amount of data. Since this involves timer-based measurements, values need to be compared with some appropriate margins. The QML documentation of this API is omitted on purpose; the QAudioFormat type is not available in QML, and we don't want to encourage users to operate on raw bytes from QML anyway. [ChangeLog][QtTextToSpeech][QTextToSpeech] Added the ability to produce PCM data as a QByteArray. The QtTextToSpeech module now depends on QtMultimedia on all platforms. Fixes: QTBUG-109837 Change-Id: I308a3e18998827089c0f75789b720f1bd36e3c46 Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Emit information about speech progressVolker Hilsheimer2023-02-182-0/+134
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This is useful information in a UI that wants to visualize the progress by highlighting the words and sentences as they get read. For this to work, we ideally can emit data through a signal for each word that allows an application to the progress information to the text that was previously passed into QTextToSpeech::say, i.e the index and length of the word within that text. Implement this for all engines where we can, and add a test that verifies that we get correct information: On the macos and darwin backends, the delegate gets called for each word about to be spoken, with index and length of the content relative to the text. We don't get access to more detailed information, like the length of the stream in second or samples, or the current playback state. Android provides an equivalent listener callback that tells us which slice of the text is about to be spoken. In the WinRT backend, we can ask the speech synthesizer to generate track data for the generated audio, which gives us access for each sentence and word, with the start time for each. Since we play the PCM data ourselves, we don't get called with progress updates, but we can use the track information to run a timer that iterates over the boundaries with each tick. This has a risk of getting out of sync with the actual playback though, but we can try to compensate for that. We can use a similar strategy on flite, where the symbol tree provides start times for each token. So we can use a timer, and follow the progress through the input text for each token. On speechd we don't have reliable access to anything; it theoretically supports reporting of embedded <mark> tags when the input is SSML. So for now, speechd cannot support this functionality. Add highlighting of the spoken word to the Qt Quick example. Change-Id: I36ff208b2f0112c9eb261864515ba20c4bf55f25 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io> Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
* Add QTextToSpeech::engineCapabilitiesVolker Hilsheimer2023-02-071-0/+22
| | | | | | | | | | Plugins can register through the plugin json file which capabilities the engine supports, which then allows applications to check what QTextToSpeech APIs they can use. Change-Id: Id22ac55a3731591ed8bb53e8db76705de10e814f Reviewed-by: Jeremy Whiting <jpwhiting@kde.org> Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Fix QSKIP usage in pauseResumeVolker Hilsheimer2023-01-201-1/+1
| | | | | | | | | The macro is declared as variadic, but in practice takes a only single parameter with the skip-reason. Pick-to: 6.5 6.4 Change-Id: Ica0f9dfcf94e09b0e15745313285c3b5ac89f8e7 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Blacklist tst_QTextToSpeech::pauseResume(darwin) on macOS 13 in CITor Arne Vestbø2022-11-041-0/+2
| | | | | | | | | It's flakey/failing, but can not be reproduced locally. Task-number: QTBUG-108205 Pick-to: 6.4 Change-Id: I4b17ca5570fbb4f2c4bf683c02dca4beb6c5108c Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Port from container::count() and length() to size()Marc Mutz2022-10-082-32/+32
| | | | | | | | | | | | | | | | | | | | | | This is a semantic patch using ClangTidyTransformator as in qtbase/df9d882d41b741fef7c5beeddb0abe9d904443d8: auto QtContainerClass = anyOf( expr(hasType(cxxRecordDecl(isSameOrDerivedFrom(hasAnyName(classes))))).bind(o), expr(hasType(namedDecl(hasAnyName(<classes>)))).bind(o)); makeRule(cxxMemberCallExpr(on(QtContainerClass), callee(cxxMethodDecl(hasAnyName({"count", "length"), parameterCountIs(0))))), changeTo(cat(access(o, cat("size"), "()"))), cat("use 'size()' instead of 'count()/length()'")) a.k.a qt-port-to-std-compatible-api with config Scope: 'Container', with the extended set of container classes recognized. Change-Id: Ib7ee4af5785944e388b550285b7d67a585c38468 Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Jeremy Whiting <jpwhiting@kde.org> Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Detect speechd misconfigurations in tst_QTextToSpeechAxel Spoerl2022-09-161-2/+9
| | | | | | | | | | | | | | | | | | | | | | | | init() used QTRY_VERIFY to assert that the speechd engine reported ready after construction. The following QSKIP statement was not reached if the QTRY_VERIFY failed. This patch removes the QTRY_VERIFY, which leads to speechd being skipped, when an error is reported after engine construction. sayWithVoices created a test text for each available voice to be spoken. The default speech-dispatcher installation on RHEL 9 provides 158 voices, which leads to the test failing with a timeout. This patch ends the loop with a qWarning after 10 voices. Timeout is thereby prevented, if any engine has too many. Both fixes are combined in one commit, because they would fail CI if staged separately. Fixes: QTBUG-106286 Pick-to: 6.4 Change-Id: I8823bbf567c47229966041611f1defb3f75f9fc6 Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Change the license of all CMakeLists.txt and *.cmake files to BSDLucie Gérard2022-08-243-3/+3
| | | | | | | Task-number: QTBUG-105718 Change-Id: I8ddfa6b3741acb3953446ef48d80bf5dde9f828d Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io> Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
* CMake: Don't use PUBLIC_LIBRARIES for tests and test helpersAlexandru Croitor2022-07-282-4/+4
| | | | | Change-Id: I203344f2e0c1c0335e8c98fef811a679f68eda1d Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Add license headers to cmake filesLucie Gérard2022-07-073-0/+9
| | | | | | | | | | | | | CMakeLists.txt and .cmake files of significant size (more than 2 lines according to our check in tst_license.pl) now have the copyright and license header. Existing copyright statements remain intact Task-number: QTBUG-88621 Change-Id: I947479e0fc6301e1622478a2c5b41a269dc8407e Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
* Replace blacklisting of tests with QEXPECT_FAIL and QSKIPVolker Hilsheimer2022-06-222-16/+34
| | | | | | | | | | | | | | | | | We know that certain features and voices are broken with certain engines due to bugs in the native APIs, and we can't work around those. Try to avoid dysfunctional voices for speechd, and document bugs in native APIs in the test code via QEXPECT_FAIL and QSKIP. Remove BLACKLIST file. Fixes: QTBUG-55274 Pick-to: 6.4 Change-Id: Ib8f26e60346ac7ca95f60433ea7e95879d0b0422 Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Axel Spoerl <axel.spoerl@qt.io> Reviewed-by: Jeremy Whiting <jpwhiting@kde.org> Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Stabilize sayWithRates testVolker Hilsheimer2022-06-221-11/+12
| | | | | | | | | | | Only start measuring time when the engine starts speaking, and move the connection and lambda outside the for-loop so that connections don't accumulate. Pick-to: 6.4 Change-Id: I6f0009ea6101a8c57c1d04e1a928d9f73296c5dc Reviewed-by: Jeremy Whiting <jpwhiting@kde.org> Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Use SPDX license identifiersLucie Gérard2022-06-222-70/+4
| | | | | | | | | | | Replace the current license disclaimer in files by a SPDX-License-Identifier. License files are organized under LICENSES directory. Pick-to: 6.4 Task-number: QTBUG-67283 Change-Id: I5a15004abaab3f2d002adf47ae053b95abb41cb8 Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
* Rename iOS engine to DarwinVolker Hilsheimer2022-06-132-2/+2
| | | | | | | | It's the default engine for both iOS and macOS. Pick-to: 6.4 Change-Id: I02c9ac77f10ad6faccf32e6ed58399890cfc770c Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* Improve error handling in QTextToSpeechAxel Spoerl2022-05-202-4/+4
| | | | | | | | | | | | | | | | | | Add an enum class ErrorReason for error handling in QTextToSpeech. Rename BackendError to Error in enum QTextToSpeech::State, as the error might not be in the synthesizer, but the audio playback, or due to invalid input. Update all sources and tests referring to BackendError to use new name. Update documentation of enum QTextToSpeech::State, add documentation for new enum QTextToSpeech::ErrorReason. Add getters, setters, signals, and slots for error handling. Implement error handling for flite plugin, covering audio backend, initialization and synthesizing errors. For the other plugins, handle basic initialization errors. Change-Id: I233cf3876511176dd1a327546233d527596e1e7e Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Simplify time measurement in sayWithRates testVolker Hilsheimer2022-05-061-4/+6
| | | | | | | | Measure the time immediately in a lambda connected to the signal, rather than wait for the QSignalSpy::wait function to return. Change-Id: Ibbeb9ad310dc3319f0e58b37bc13f51b33d2595b Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* Check audio device availability only if the engine is not mockVolker Hilsheimer2022-04-291-4/+4
| | | | | | | Qt Multimedia qFatal's when there is no backend, crashing the test. Change-Id: Ib7d4db521a32e57dc00be6339501dd7b37b4a2e3 Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Workaround for iOS engine on macOS 10.14Volker Hilsheimer2022-04-292-0/+12
| | | | | | | | | | | | | The documentation claims that the AVFoundation based frameworks are available from macOS 10.14 onwards, but the API for getting the voices returns nothing. Handle that in the test case for now. Ideally, our plugin infrastructure would allow plugins to report such conditions so that the respective engine gets ignored. Change-Id: I0d7ab79acf41a9aea752e1789a74209faf689d3c Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Make test robust on systems with a single localeVolker Hilsheimer2022-04-291-1/+3
| | | | | | Change-Id: Ife4e719899a84482ab1f0e95028081f51f737b7f Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Skip tests with old libspeechd version (pre-0.9)Volker Hilsheimer2022-04-282-14/+11
| | | | | | | We don't want to support it anyway. CI provisioning needs to be fixed. Change-Id: I897f7c40178a35bb096ea470942ff54d6204756e Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Log more information in case of test failureVolker Hilsheimer2022-04-281-1/+4
| | | | | | | | | The test failed with the mock engine on macOS on ARM. The mock engine uses a QTimer that waits for a rate-dependent time on each word before switching back to Ready state, so that should not be possible. Change-Id: Id54bcdc6fa19b86d93208724fc839d81a95e0f4a Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Blacklist speechd tests in CIVolker Hilsheimer2022-04-281-1/+9
| | | | | | | | | On OpenSUSE, an old libspeechd is installed that we no longer want to support. On Ubuntu, a newer version is installed but with broken default voices from the mary-generic festival module. Change-Id: Ifde0fcd7378313aa7817654d57c2f08796ab81be Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Blacklist pauseResume test with macOS engineVolker Hilsheimer2022-04-281-0/+2
| | | | | | | | The NSSpeechSynthesizer functionality for that is broken, and we'll remove that engine entirely. Change-Id: I945c98961cc68db56466aff41927f07e6260cdea Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* Skip speaking tests if there is no audio deviceVolker Hilsheimer2022-04-272-0/+26
| | | | | | | | | | | | Those tests will fail, as the audio playback will either not start at all or finish immediately. Skip those tests, unless we use the mock engine. Fixes: QTBUG-82545 Change-Id: Ib3b51a1b38928bcbea20f1515d231637608fd206 Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io> Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
* Add QVoice::locale propertyVolker Hilsheimer2022-04-201-0/+1
| | | | | | | | | | | | In every engine, the locale is a property of the voice, so store it in the voice. This makes sense for applications that want to know which language a voice speaks, and it allows us to simplify the code for engines that used to store the current voice and locale: one is enough. Change-Id: I9a01ebf2d5817accf820ac5dee6a7f24524195d7 Reviewed-by: Axel Spoerl <axel.spoerl@qt.io> Reviewed-by: Jeremy Whiting <jpwhiting@kde.org> Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* Android specific build and test fixesVolker Hilsheimer2022-04-141-7/+19
| | | | | | | | | | | | | | | | | | | | | | | | | | Make the plugin depend on the jar target, otherwise the jar does not get built when needed. And since the Android engine initializes asynchronously we need to always wait for the Ready state. This makes the tests for properties pass; the tests waiting for the speaking to finish abort the test run. Fix the implementation of pitch and rate conversions to correctly map from Qt's ranges of [-1.0, 1.0] to Android ranges of (0.0, 2.0[, and [0.5, 2.0], respectively. Since Android operates on floats a rate of e.g. 0.9 becomes a double of 0.89999 and the test fails, so use string-based testing which allows us to limit the precision. Add a "warmup" utterance to the sayWithRates test so that the first utterance doesn't take longer due to lazy initialization. With these changes, the auto test now passes on Android. Change-Id: Ie7698bcfdd60348441b1ea2d53cd2126d4355d5d Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* Add tests for setting voice and localeVolker Hilsheimer2022-04-141-0/+119
| | | | | | | | | | | | | | | | | The two properties are intertwined, changing one is very likely to change the other: setting a locale in which the current voice is not supported has to change voice; and setting a voice that the current locale does not support has to change locale. Emit signals if, and only if, the respective other attribute has been modified after calling the engine on the setter. The engine's setter implementation is responsible for picking a new voice or locale, and for testing that the setter changed the value in the first place. Adjust mock, flite, and speechd engines accordingly. Change-Id: Iafd585b56a1a3c36b919ed6e6f29aabbd322e97c Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
* Don't emit signals when attributes don't changeVolker Hilsheimer2022-04-141-0/+8
| | | | | | | | | | | | | Leaving it to the engine implementations to ignore and return false in setters when the new value is the same as the old value is error prone and unnecessary, even if engines can sometimes chose a slightly more optimal way to compare values. Ensure that we have the right semantics in the QTextTosSpeech class directly, and add test coverage. Change-Id: Ie18b9bf4577e18f9fca9f70d1310856d1ab35f7b Reviewed-by: Jarkko Koivikko <jarkko.koivikko@code-q.fi>
generated by cgit v1.2.3 (git 2.25.1) at 2024-05-24 23:27:06 +0000