I'm happy to test on Windows 7, and on Raspbian Linux (Although Raspbian Linux runs Pharo UI events slo-o-owly, so I am not the best person to do the Linux testing) On 12 November 2015 at 15:42, EuanM <euanmee@gmail.com> wrote:
I think anything that makes a consistent UI to Pharo for users on different platforms is a good thing.
However, I do not know what the downside-effects of this change are.
The key thing is to capture all breaks caused by the changed, so they can be coded through or around. And ensure that the test results are viewed with at least the same importance as the automated Jenkins CI test results.
I do not know the extent that the Jenkins CI tests actually simulate user keyboard input, so... (Pharo pseudocode for comic effect and clarity)
We can write Jenkins tests that simulate user input ifTrue: [ we should do so ]. ifFalse: [ we should write a canonical set of regression test scripts for human users to follow . we should do: [:eachPlatformSupportedByPharo | run the tests in which humans use keyboard tasks ] . we should capture the results ; and feed them into automated the regression tests results database ]
On 12 November 2015 at 08:35, Nicolai Hess <nicolaihess@gmail.com> wrote:
A short overview:
With Ctrl-Key pressed,
some keyboard events don't generate smalltalk keystrokes (EventKeyChar) ctrl+tab, ctrl+1, ctrl+2, ..., ctrl+0, ctrl+Tab, ctrl+m
for some keyboard events there are only key down and key up events (from windows) but the vm "generates" a keypress (EventKeyChar) event ctrl+Home, ctrl+End, ctrl+PageUp, ...
But the key value , char code and ctrl flag for this generated events, make them undistinguishable from other ctrl+key shortcuts ctrl+a = ctrl+Home, ctrl+d = ctrl+End, ctrl+k=ctrl+PageUp,...
And one key (ctrl+Enter) creates two EventKeyChar events one "normal" and one that is generated from the vm one with keyvalue 10 and one with 13.
And the keycode for char like "a","b", ..., are different if you they are typed with or without ctrl-key
just key "a": keycode: 65 (keydown event) 97 (keyvalue/charcode event) (65 if you press shift key) 65 (keyup event)
with ctrl key pressed: keycode 65 (keydown event) 1 (keyvalue/charcode event) 65 (keyup event)
I would like to change this, that means:
generate EventKeyChar events for ctrl+1, ctrl+2, ..., ctrl+Tab, ctrl+m
remove the double key event for ctrl+Enter
and adopt the behavior of the linux vm, for keyValue, charCode values for events like ctrl+a, ctrl+home, ...
for example, the linux vm generates event buffer values: #(2 7329656 1 0 2 65 0 1) for ctrl+a (keyValue 1/charCode 65) but #(2 7329656 1 0 2 1 0 1) for ctrl+home (keyValue 1/charCode 1)
this may need some changes on the image side. For example, in Pharo we have some special handling that adds "96" to the keyValue for keystrokes with ctrl-key).
What do you think?
nicolai