Smaller bug fixes in the editor libraries#8324
Conversation
…e the repaint area on 2x HiDPI scaling.
The bug was easy to reproduce in the IDE: 1) Create an empty text file 2) Enter one line of text 3) Move the cursor to the start 4) Press Enter a few times. Frequently, when the necessary repaint isn't coincidentally triggered by the cursor blinking, the line will appear painted multiple times (because the previous position of the shifted line isn't repainted).
…ing off the paintable edge at the end of a long line.
matthiasblaesing
left a comment
There was a problem hiding this comment.
Did a quick test on a "normal" screen. Caret still seems to work correctly, I reproduced the redrawing problem and found it fixed with this change. Eyeballed the rest and looks sane to me.
|
Thanks for testing and reviewing! |
|
When I press "Merge pull request" I get a new prompt now that I haven't seen before: 
Is this the right way to merge? (In this particular PR's case I do want to keep the commits separate.) |
|
Just looked into another project and there all options (plain merge, squash and merge and rebase and merge) are available and for merge it looks identical. The GUI makes sense when you come from git CLI client as for a non-ff merge a commit message is required there too. As the email is selectable, I assume that it will be correctly recorded. |
|
Thanks! Yeah, I merged and it looks the same as other merged PRs. |
Here are a few smaller bug fixes in the ide/editor.lib2 module, each in a separate commit (since they are independent of each other).
For platform apps that use the THIN_LINE_CARET option, on HiDPI monitors, fix a paint bug where a pixel-thick line from the caret may remain outside the repaint area, leaving "cursor droppings" behind as the user types.
Avoid a NullPointerException that could occur if an editor is closed while MacOS trackpad scrolling events are still coming in.
Fix an editor repaint bug which could leave lines appearing duplicated. Easy to reproduce on plain text files; see the commit for reproduction steps.
The cause was clear: An existing comment in o.n.modules.editor.lib2.view.ViewBuilder explains that "For valid firstReplace the current impl repaints whole line", but that repaint was in a block of code that was being skipped as an optimization (the "accurate span" calculation). The fix is to check for the "firstReplace" case and ensure that repaint() will be called regardless (via the checkLayoutUpdate method).
When line wrapping is enabled, adjust the wrap line length so that the text caret can't end up completely outside the scroll viewport (making it invisible).