libiconv status report #4

[Gabor Kovesdan]

Hi,

this week's progress:
- Investigated about UTF-8 and found out that the differences between 
GNU iconv and Citrus are bugs in GNU because it allows unpaired 
surrogates in the output, which shouldn't be allowed according to 
unicode.org.
- Also checked UTF-7 and the behaviour seems to be completely the same 
apart from that Citrus gives some warnings on invalid sequences, which 
is fine I think.
- Added mapper_parallel modul, which wasn't built before
- Little optimization to UTF-8: static tail length lookup tables instead 
of filling the table with for loops.
- Fix Kazakh conversion table paths, so Kazakh conversions work now.

I have some more changes in progress but they aren't ready yet, so I 
haven't merged them. I consulted with my mentor and I'll organize my 
work around GNU iconv compatibility so that we can start to use iconv as 
soon as possible. Optimization can be done when I find something like 
this UTF-8 nit and I'll review the whole code at the end of the program. 
My mentor also pointed out that glibc iconv seems to be even better than 
GNU iconv, so I'll take a look at it, as well. From the next week I'll 
have limited connectivity for some time but as I do weekly merges, you 
might not notice it at all.

Cheers,

-- 
Gabor Kovesdan
FreeBSD Volunteer

EMAIL: gabor at FreeBSD.org .:|:. gabor at kovesdan.org
WEB:   http://people.FreeBSD.org/~gabor .:|:. http://kovesdan.org