On Fri, Sep 19, 2014 at 04:55:24PM +0900, Daiki Ueno wrote:
> Ben Pfaff <b...@cs.stanford.edu> writes:
>
> > New module 'unistr/u8-mb-prev-uc'.
> > New module 'unistr/u16-mb-prev-uc'.
> > New module 'unistr/u32-mb-prev-uc'.
>
> Thanks, looks good to me. Some nit-picking below.
>
> * _GL_UNUSED_PARAMETER of u32_mb_prev_uc seems to be a leftover, as the
> argument is actually used in the implementation.
>
> * The largest value of possible leading octet tested in
> test-u8-mb-prev-uc.c:exhaustive_test is 0xf5, while there are checks
> against 0xf8 in u8-mb-prev-uc.c. Also code units above the surrogate
> code-point are not checked in u16 and u32 tests.
> * The license of u{8,16,32}-mbtouc changed to LGPLv2+ some
> time ago, maybe good to follow the change?
All of the above are good points. I will fix them in v5.
> > v3->v4: Changed the code to always be "safe". It looks to me like the
> > "unsafe" version that I had written originally reflected a
> > misunderstanding of how the gnulib option for that was supposed to work.
>
> Are you going to add "unsafe" version later, or is it not useful at all?
I do not have a use myself for unsafe versions, but I will add them if
you think they are a good idea. Let me know.
