In CreateAndSetFocus(), SetInputScope() is called *after* setting focus to the context. At this time, Google Japanese Input retrieves InputScopes. Therefore, TSFTextStore returns IS_DEFAULT. But after that, Google Japanese Input tries to retrieve InputScopes after every notification (in this case, a call of ITextStoreACPSink::OnLayoutChange()). Then, we return IS_URL due to set after returns IS_DEFAULT.
This is actually our fault, but according to the other TIPs, Google Japanese Input shouldn't commit composition at detecting an InputScope change, though.
MozReview-Commit-ID: 2tPlcEA0MI0
I'm still not sure what we should do in this case, though.
If mContentForTSF is initialized and there are some unknown changes in actual contents, i.e., not caused by composition of the active TIP itself, we cannot set selection range properly in some cases.
For example, if TSF tires to set non-empty selection range but the range has been removed by web apps.
For now, let's try to return E_FAIL in such case because that should occur at reconversion or something immediately after previous content change not caused by previous composition. If TIP does nothing in this case, user can retry with same operation after all pending text changes are notified to TSF.
MozReview-Commit-ID: 9unrNVeC1tW
Same as selection change notification, text change notification shouldn't be notified to TSF while there is cachec content because neither TSF nor TIP may allow to change text by web applications during keeping storing cached content.
This patch makes TSFTextStore stores and merges text changes until MaybeFlushPendingNotifications() is called and there is no cached content.
MozReview-Commit-ID: 9fj0GREbX18
TSFTextStore shouldn't notify TSF of selection change until MaybeFlushPendingNotifications() is called and there is no cached content because while there is cached content, neither TSF nor TIP may allow to change selection by web applications. Therefore, ITextStoreACP::GetSelection() and similar methods need to use mSelection instead of actual selection in the focused editor. Therefore, TSFTextStore should store selection change data during keeping storing content cache and notify it when the cache is cleared. So, when TSFTextStore notifies TSF of selection change, TSFTextStore needs to update mSelection to the actual selection which is stored in mPendingSelectionChangeData.
MozReview-Commit-ID: 8ZWASzu7Znv
This patch stop clearing mContentForTSF at unlocking the document because we should keep it until active composition is committed. If so, TSF/TIP won't be confused by content changes by JS. So, this is important for a11y of TIP users in some complicated websites like GoogleDocs, Facebook, etc.
Note that this patch doesn't work well without following patches. We need to stop notifying TSF of selection changes and text changed while mContentForTSF is valid.
MozReview-Commit-ID: 9QOGZxdYU3I
I think that we can drop nsIMEUpdatePreference::DontNotifyChangesCausedByComposition(), i.e., nsIMEUpdatePreference::NOTIFY_CHANGES_CAUSED_BY_COMPOSITION because it's now used only by TSFTextStore but TSFTextStore ignores if SelectionChangeDataBase::mCausedByComposition or TextChangeDataBase::mCausedOnlyByComposition is true (for supporting async changes in e10s mode). So, only issue is, dropping the flag might cause increasing computing TextChangeData cost during composition in TSF mode. However, now, it's already enough fast and even if it'd cause performance regression, we could add a hack with TextComposition's offset information. Therefore, we don't need to worry about the performance regression so seriously.
MozReview-Commit-ID: HNT3G4isONj
Currently, all widgets request selection change notifications to IMEContentObserver. Additionally, IMEContentObserver needs to listen selection changes for caching latest selection for eQuerySelectedText. Therefore, it doesn't make sense to keep defining nsIMEUpdatePreference::NOTIFY_SELECTION_CHANGE.
If widgets didn't need selection change notifications, they could just ignore the unnecessary notifications.
Note that all widgets don't need selection change notifications if a plugin has focus and IMEContentObserver cannot observe selection changes in the plugin. Therefore, if IMEContentObserver is initialized with a plugin, it shouldn't listen selection changes (and doesn't need to notify widgets of selection changes).
MozReview-Commit-ID: FOVFFgA2nOz
TSFTextStore should use insertion point relative query for getting character rect and caret rect while there is a composition because in such case, TSF must want to query a part of or around the composition. Therefore, it makes sense to use insertion point relative query since it adjusts the offset with the latest content information.
MozReview-Commit-ID: IVjZ4zqFUkr
While a TSFTextStore instance is being destroyed, TSFTextStore::Destroy() tries to commit remaining composition and notify TSF of destroying the view. At this moment, TSF/TIP may try to commit the composition or retrieve the contents with calling ITextStoreACP::RequestLock() but currently TSFTextStore disallows the requests to lock of them. This means that TSFTextStore never sends composition commit events asynchronously. Therefore, TextComposition may keep waiting remaining composition events but this causes odd behavior because they won't be fired.
For avoiding this issue, TSFTextStore should behave as normal even while it's being destroyed. Fortunately, if there is a composition, it always has mLockedContent and mSelection. So, it can compute expected results of TSF/TIP with them.
MozReview-Commit-ID: 2DSCGXXkLx1
While TIP is handling a key message, TSFTextStore shouldn't release any TSF objects since it may cause hitting a bug of TIPs. Actually, MS-IME for Japanese on Windows 10 crashes when TSFTextStore is destroyed during composition because probably it accesses some destroyed objects to request to commit composition or query contents.
MozReview-Commit-ID: 9CTjHhAvG04
For making our code clearer by the stronger type check, we should change the anonymous enum for NS_TEXTRANGE_* to enum class whose name is "TextRangeType" and whose type is "RawTextRangeType" which is an alias of uint8_t.
Additionally, this also adds some utility methods for them.
Note that some lines which are changed by this patch become over 80 characters but it will be fixed by the following patches.
MozReview-Commit-ID: 76izA1WqTkp
It's not clear to me what NOTIFY_IME_OF_COMPOSITION_UPDATE means only from the name. For making the name clearer, this patch renames it to NOTIFY_IME_OF_COMPOSITION_EVENT_HANDLED and add some explanation to the definition.
MozReview-Commit-ID: 8ySYCNJ1Ytz
We won't enable TSF mode on Windows XP nor Windows Server 2003 because they are already not supported by Microsoft itsef and installed TSF of them is too old and too buggy. Therefore, it's not worthwhile to support TSF on those environments anymore.
MozReview-Commit-ID: Cs85EvWie9K
TSFTextStore::sEnabledTextStore is a static variable to grab a reference to focused TextStore instance. So, this may be changed by accidentally during a call of instance methods of TSFTextStore. Then, focused TextStore may be destroyed during running a method and crash when it accesses a member variable.
For avoiding this crash, static methods which call a method of sEnabledTextStore should create an independent RefPtr to it before calling the method.
MozReview-Commit-ID: 51Sor1LdABr
