The platform window is the virtual screen for Cappuccino windows. Per Cocoa, when the screen (platform window) resizes, CPApplicationDidChangeScreenParametersNotification is sent and the CPApplication delegate method applicationDidChangeScreenParameters: is called.
Previously, window moves were constrained such that a minimum margin of the window was visible on screen, but window resizes were not constrained, allowing the user to effectively move the window off screen.
Now resizes are constrained such that a minimum margin of the window is always visible on screen so that the window does not get "lost".
Previously, whenever the platform window (virtual screen) resized, windows were moved or resized to keep them on screen. This could not be easily overridden by the user.
Two class methods have been added:
CPWindow +setConstrainWindowsToUsableScreen:
CPWindow +constrainWindowsToUsableScreen
If +setConstrainWindowsToUsableScreen:NO is called, then resizing the platform window will not move or resize user windows.
Previously, _CPAutocompleteMenu was trying to relayout itself using its related textfield's window frame. This was
causing a crash if one of the CPTokenField's superview has been removed. This fix ensures to stop layouting if the related text field has no window anymore.
Previously, in IE you could scroll the entire browser window to the right a very long way, leaving white space to the right of the Cappuccino app. This was caused by a 100,000px wide measurement span that was supposed to be positioned at left:-100,000 but was mistakenly at left:-10,000, which left 90,000px on screen.
This commit fixes the left position of that span, eliminating the ability to scroll the app.
Fixes#1725
Previously, unless you were using a sheet alert, there was no way to set a custom dismissal handler for an alert. In Cocoa alerts are synchronous, in Cappuccino they are asynchronous, which meant that a single dismissal handler had to handle all possible non-sheet alerts.
This commit adds several methods which add two ways to specify a custom dismissal handler: via delegate/selector, and via a block (function). The new methods are:
setDidEndSelector:
runModalWithDidEndBlock:
beginSheetModalForWindow:didEndBlock:
Note that didEnd blocks are transient and take precedence over didEnd selectors.
The CPAlertTest demo was updated to fix some bugs and added the ability to run the tests using blocks.
Previously the inner element's offsetWidth was tested before and after setting the overflow to scroll. This is unnecessary, we can just set the overflow to scroll and test the outer element's clientWidth vs. offsetWidth. By definition, clientWidth does not include scrollbars, whereas offsetWidth does. So if they are equal then overlay scrollers must be in use.
Note that FireFox (as of version 22) does not support overlay scrollers, so even if the system does, Cappuccino will not on FireFox unless the programmer forces overlay scrollers.
Previously, if the last row's height was increased, empty rows below would not be pushed down. This was due to an off-by-one in -_unboundedRowsInRect (now _exposedRowsInRect) which caused the incorrect rect of the last row to be retrieved.
This commit fixes that bug and also makes the following changes:
- Added FULL_ROW_HEIGHT(), ROW_BOTTOM() and HAS_VARIABLE_ROW_HEIGHTS() macros to make the code clearer.
- Documentation cleanup.
- Eliminated some intermediate variables.
- Rewrote -_rectOfRow to reduce redundant code and fixed some logic errors.
- Renamed some variables to make their intention clearer (to me at least).
- Added an out of bounds check in -rowAtPoint: before doing the relatively expensive binary search.
- In -noteHeightOfRowsWithIndexesChanged:, the first valid index is used instead of the first index (which might be invalid).
- Fixed a bug in -noteHeightOfRowsWithIndexesChanged: where row height was not set correctly if a row was not in the index set.
- Fixed incorrect use of CEIL in -_unboundedRowsInRect, it was wrapped around the divisor instead of the entire expression.
- Fixed drawBackgroundInClipRect: drawing one too many rows.
- Fixed -rowAtPoint not checking for x in bounds.
BREAKING CHANGE:
Previously, CPTableView -rowAtPoint would incorrectly return a valid row index when the x value was outside of the table's bounds. Now an out of bounds x value will return -1, which is consistent with Cocoa (and the intent of the method).
Previously, the auto-enabling logic for CPMenu would not check to see if a menu item had enabled bindings, and would ignore the state determined by those bindings.
Now, if a menu is auto-enabling and an item has an enabled binding, the binding is used to set the enabled state of the item.
Previously setObjectValue checked for a formatted value, and then called [self stringValue], which ended up formatting the value again.
Now all of the cases (no formatter, formatter succeeds, formatter fails) are handled explicitly in setObjectValue, eliminating the extra call to the formatter.
Previously, no attempt was made to become/resign first key responder when a CPTextField was unhidden/hidden. This could lead to a case where a hidden CPTextField was made first responder, then unhidden (and have a focus ring), but would not receive key input.
Cocoa specifies that viewDidHide will be called when an unhidden view is added to a view hierarchy with a hidden ancestor view, and viewDidUnhide is called when an unhidden view is removed from a view hierarchy with a hidden ancestor view. This was not the case in Cappuccino.
This commit adds that functionality, and also removes a spurious viewDidHide/viewDidUnhide call that was generated as a side effect in initWithCoder.
Previously, _createDOMImageElement was not called until after [super initWithCoder] was called. But that call would end up in CPImageView -setObjectValue, which would attempt to access _DOMImageElement before it had been created.
Now _DOMImageElement is checked before being accessed or created.
Previously, a case could arise where a text field was set as the initial first responder but during cib loading was not yet assigned to a window. In this case _isWithinUsablePlatformRect would fail.
Now _isWithinUsablePlatformRect checks to make sure the text field has a window.
Tagging with UIDs can make it much easier to debug encoding/decoding related bugs where it's not clear that the right view ended up in the right spot.
The new data attribute is called "data-cappuccino-uid".
Also collect and simplify appkit_tag_dom_elements code a little.
Without this fix, the `appkit_tag_dom_elements` flag would only affect CPImageViews unarchived from a coder (e.g. in a collection view, from a cib). It would have no effect on a CPImageView created in code.
This change refactors the _DOMImageElement code to a shared method, reducing redundancy while fixing the bug.
Without this fix, code like `[[[CPImage alloc] initWithContentsOfFile:nil] description]` would cause a crash.
This fix makes `[[CPImage alloc] initWithContentsOfFile:nil]` behave like its equivalent in Cocoa: it returns nil without warning or error.
If a cib file had a custom CPWindow subclass, that cib file would not be properly processed by flatten as it'd fail to find the relevant class (the fact that the class can't be found is an issue in itself). This would cause an exception.
This fix makes CPCibWindowTemplate act similar to CPCibCustomView when the referenced class can't be found: it instantiates a regular CPWindow. This is sufficient for flatten's purposes.
Previously, certain situations could easily lead to an infinite loop in - CPView nextValidKeyView. For instance, opening a sheet view with no responder returning YES for canBecomeKeyView would freeze Cappuccino.
This was caused by the cycle detection in nextValidKeyView not being able to pick up on cycles where the nextKeyView of the original receiver did not participate. Since content views often have a next key view inside of them, and the views inside rarely have a next key view back "up" to the content view, the content view would never be seen again even if there was a cycle.
This fix detects cycles at any level by remembering all previous visited responders.
This change brings nearly all copy and paste related into a single, self-contained class. This new class is still somewhat coupled to CPPlatformWindow+DOM.j in the sense that it's highly specific and meant to be called in a very specialised way. Despite this, it's still a significant decoupling and the new code is more readable and easier to follow.
Without this fix, if a menu item was disabled it could only be automatically re-enabled if the menu item had a `validateMenuItem:` or `validateUserInterfaceItem:` enabled target.
This meant that for example if the Edit > Select All menu item was validated when the first responder was something like a collection view, it'd become disabled. If then the first responder was changed to a text field and revalidation occurred, the menu item would not become enabled and Select All would not be possible in the text field neither through the Edit menu nor the Cmd-A/Ctrl-A keyboard equivalent.
This fix ensures that menu item validation does not only take negative action (disabling enabled items which should be disabled), but also positive action (enabling disabled items which should be enabled), even when there's no `validateMenuItem:` or `validateUserInterfaceItem:`.
This change begins to clean up the control flow of copy and paste. Before, copy and paste was dealt with partly in CPPlatformWindow, partly in CPApp and partly in CPTextField.
Now, CPApplication has been removed from the path, simplifying it down to the two more expected actors.
These targets might be rare but could occur in some Cappuccino text editor widget, in which case our standard native paste handling from Safari should work fine.
Without this fix, when text was copied from a CPTextField, the value would be copied into the system clipboard (if possible), but not into the Cappuccino pasteboard. This meant that the text could not be pasted back in Cappuccino using e.g. Edit > Paste from the Cappuccino menu.
This fix ensures the appropriate text is copied both into Cappuccino's pasteboard and the system clipboard (when possible).
This fix switches to full native copy and paste in browsers which support it. This allows copy from anywhere - a collection view, a table view and so on - without the browser greying out the copy menu or beeping.
In the future this can easily be extended to copy and paste images as well.
This change also improves Safari paste support. In recent versions of Safari, pasting from the system clipboard into Cappuccino has been impossible. This remains impossible due to the lack of beforepaste events, but now the code cleanly switches to a Cappuccino-only paste at least.
Without this feature, attempts to paste data into a Cappuccino app (outside of a text field) are handled by redirecting the paste into a hidden text field. Then a paste event is generated from whatever is captured in this field after a 0 timeout. This method is prone to timing related bugs where the paste event actually has an empty clipboard.
With this new feature we read the paste data directly from the browser's native paste event when the browser supports it (Chrome 10+, Safari 5+, Firefox 22+). This is much more dependable and may in the future also enable us to read other types of paste data such as images.
Firefox is not accepting the initial values for these two attributes, making it crash with an ILLEGAL VALUE error in any application using CPBezierPath (especially ones with CPPredicateEditor that is using bezier path to draw the selection arrows.)
Aparajita Fishman
Antoine Mercadal
Martin Carlberg
Alexander Ljungberg