I'm fixing an extremely "worthy" bug

[Edward K. Ream]

PR #4624 created an extremely nasty bug, one of the "worthiest" in Leo's history. Happily, this bug happened after releasing Leo 6.8.8, so the pain is limited to those testing the devel branch.

After several false starts, I have chosen to start over from the  last good commit reported by git bisect. There was no other obvious way forward.  There was no joy after several trial git merges!

PR #4631 contains the new work. Phase one fixes the bug. Phase two will restore work clobbered by taking six entire files from the last good commit. 

Please hold off further related bug reports until  PR #4631 is ready for review. Thanks.

Edward

[Edward K. Ream]

On Wednesday, April 22, 2026 at 2:23:33 PM UTC-5 Edward K. Ream wrote:

> PR #4624 created an extremely nasty bug...

> PR #4631 contains the new work.

> Phase one fixes the bug. Phase two will restore work clobbered by taking six entire files from the last good commit. 

Work is going well. At first I felt daunted by the tasks. How could I ever make sense of huge diffs? Happily, the job is almost entirely about process--running tests over and over again. That's a comforting realization.

This morning marked a milestone: I created a unit test, running under the Qt gui, that ensures that #4626 won't ever happen again. It was more straightforward than I expected.

Another milestone. For ages I've been a bit afraid of git merge. I finally (Doh!) thought to google about dry-running git merge.

The --no-commit --no-ff command-line argument are exactly what I want. I've added them to my gm.cmd script so that I'll never do a "blind" merge again.

Summary

PR #4631 fixes the bug and adds a unit test that should ensure that the bug never happens again.

I could merge the PR now, but there's quite a bit of cleanup to do. See the PR for details.

Edward