bug#65726: 29.1.50; Crash in regexp engine

[Stefan Monnier]

> Thank you for fixing it, but now regex-emacs-tests fail with a regexp stack
> overflow.

Really?  I thought I had specifically run that test.
[ ...recompiling + rerunning the test... ]
Hmm... you're right.  Crap.

> It seems that by fixing this bug, we've unfixed the one that the
> broken code was supposed to fix.

Well, that one was less of a bug (more of a missing optimization), so
it's no reason to revert the change, but yes, not good.
Let's see if I can finagle something.

> And we should probably include a regression test for this bug as well.
> The previously mentioned
>
>   (string-match (rx (* "a") (* (* "b"))) "a")
>
> might suffice, but extending it to
>
>   (string-match (rx (* "a") (* (or "c" (* "b")))) "a")
>
> is safer in case the regexp compiler decides to simplify (* (* X)) -> (* X).

Feel free.  To me, it feels too much like testing the presence/absence
of a very specific bug (i.e. too unlikely that we'll reintroduce that
exact bug.  It's not like it was a weird case I didn't think of).


        Stefan