We've posted about this issue once before in
https://groups.google.com/g/reviewboard/c/7Djo6YOzwBg back in 2020. That thread has more information about how to reproduce the problem, with the file contents that were used. After creating a review using those file contents you should have a review with two revisions. The diff between original and revision 1 will look fine and so will the diff between original and revision 2 (first screenshot above). But the interdiff for revision 1 to revision 2 has incorrect colorization. In the second screenshot above you can see that line 2 has been changed from a blank line to "#if 0" but it isn't colored in yellow. Lines 3 and 4 from the original file have been deleted but they are not colored in red. This is dangerously misleading as it makes it look like there has just been a single comment added when in fact there were other changes. This example only shows a couple of lines but we've seen the same problem in bigger diffs where entire functions are added or deleted and they don't show up as diffs. Those get hidden in the collapsed sections of the file so even if you ignore the colors and examine the page line by line you're still not seeing everything that actually changed.
Can you successfully reproduce this?
--
James Abbatiello