Re: i18n backlog

[Paul Eggert]

On 2023-09-09 07:58, Bruno Haible wrote:
>    Currently the mbscasecmp tests test only valid input. Someone should
>    extend the unit test to cover strings with invalid input bytes. Then
>    we could see what difference exactly it makes.

If I recall our earlier discussion correctly, when given invalid input 
mbscasecmp's two implementations should agree in unibyte and UTF-8 
locales, but they can disagree in some other multi-byte encodings. As 
long as the implementations are self-consistent (mbscasecmp is a 
consistent total order, for example) I doubt whether this minor 
discrepancy will matter much. That being said, it wouldn't hurt to 
document some possible variances in areas where the behavior currently 
isn't documented. When doing that, it might also make sense to allow 
implementations other than mcel and mbiter, as SEE and MEE are not the 
only choices.