Editor does not refresh screen when deleting near EOF

[redmunds]

This does not reproduce in Sprint 21! Both Mac and Win (with slightly different recipes). Does not seem to be any file corruption -- just a refresh problem.

1. Open a longer file such as src/styles/brackets.less
2. Ctrl-End or Cmd-down to go to end of file
3. Select from EOF up to half the screen or so
4. Press Delete or Backspace

Results:
Huge block of unpainted screen at top of editor. It doesn't really matter how much is selected, but more makes it easier to see.

Mac Details:
Seems to require the EOF is part of selection

Windows Details:
EOF does not need to be part of selection. Needs to be last page of file so the text above selection fills in the page.

[dangoor]

Gave me a good excuse to try git bisect for the first time :)

7abaf6f (a CodeMirror update) is the culprit.

[njx]

I bisected in CodeMirror—looks like this is the actual commit that broke it:
codemirror/codemirror5@5062194

He revised that patch in a later commit, but it's still broken:
codemirror/codemirror5@9338ba4

[njx]

Filed as codemirror/codemirror5#1391

[njx]

...and already fixed :) Need to decide if we want to take it for this sprint.

[njx]

Per @adrocknaphobia, moving to sprint 23.

[njx]

Marking as "for release note" - @peterflynn, can you add this to the Sprint 22 release notes? Thanks.

[peterflynn]

Just realized this only repros if you have word wrap enabled. I've updated title & steps to reflect this.

[njx]

I don't think that's true--in the bug I filed in CM, it reproduced in a non-word-wrap demo before Marijn fixed it.

[peterflynn]

Yeah, weird... I couldn't repro without word wrap before, but now I can.

[peterflynn]

I do have to delete a pretty large number of lines (at least 1/3 of the screen, sometimes closer to half 1/2) before it will repro, though.

[njx]

Yup, that was one reason why we were okay deferring it to this sprint. Marijn's fix seems to fix it--we just weren't sure about the risk.

[njx]

FBNC to @redmunds now that the latest CM is merged--wanted to bounce this back to you so you can test around various cases on Win as well. Thanks.

[redmunds]

@njx I'm still seeing problem on both Mac and Win in master. Are you sure it got merged? Need to update SHA? Is there a different branch I should be testing in?

[njx]

Argh, sorry, looks like I didn't push the new SHA commit up. It's there now.

[redmunds]

Confirmed. Closing.

[njx]

Reopening. The fix for this bug in CM causes our unit tests to fail for an unknown reason (though only when run as an entire suite, which is mysterious). Reopening this bug until we can figure out what the problem is there.

[njx]

FBNC to @redmunds again--should actually be fixed without breaking other stuff now :)

[redmunds]

Confirmed again. Closing again.