Does anyone have a pointer to the mailing list post/write-up about OCaml catching a Unicode algorithm bug?

I think it was caught in one of @dbuenzli 's projects? Something about the OCaml typechecker pointing out a redundant or perhaps non-exhaustive case?

I’m looking for it as a ‘wow’ proof of OCaml/type safety benefit :slight_smile:

5 Likes

I believe you are referring to this UAX #14 for 13.0.0: LB27 first's line is obsolete from Daniel Bünzli via Unicode on 2020-03-03 (Unicode Mail List Archive) ?

Cheers,
Nicolás

3 Likes

Technically it’s not a bug, it pointed out a rule that was no longer needed in the Unicode line breaking algorithm.

Here’s my message to the unicode mailling list.

6 Likes

Thanks both! Saving for future reference :slight_smile:

Reading that message a year later I deeply regret saying “my compiler” on that list rather than “the OCaml compiler”. It does tell that there are not that much (enough ?) other compilers in my life…

But I insist this is not about a bug. A motto out of this would be “Luc Maranget (et al.) and the OCaml compiler keep your specifications clean”.

The proposed update for Unicode 14.0.0 has the fix.

4 Likes

It’s quite amazing how the rule looks almost exactly like the OCaml pattern. Domain modelling, indeed.