cvs commit: src/lib/libc/locale utf8.c

[Andrey Chernov]

On Thu, Oct 25, 2007 at 03:31:20PM -0400, Daniel Eischen wrote:
>> Although standard permit this, it will cause lots of programs to break
>> here and there and even may cause core dump. I don't think it is what
>> everybody wants. Non-inlined versions a) not used nowdays b) must be exact
>> as inlined.
> 
> If programs are incorrectly written, and they use inlines, then
> that is not our fault :-)

The amount of such programs is extremely big to ignore, even FreeBSD is 
not 100% clean, not say about ports. Usually ASCII-only people did't 
notice any side effects because typed 7-bit ASCII. It doesn't mean that 
the rest of the world don't exists.

> The problem with trying to do error checking and other

Error checking is permitted by standard too. We do that way traditionally. 
Nobody wants a core dump on typed char. Ad the PR addressed was about 
false positives, which would happens without error checking.

> esoteric things with inlines is that it exposes our private
> parts to the world.  You can get arrested for that ;-)

There is enough parts already exposed, like __mb_cur_max. Why you don't 
complain? 
The goal here was to keep compatibility with old binaries, so why this way 
is choosed.

> How much juice are we really trying to squeeze out of
> these inlines anyway?

Not sure what you say about.

-- 
http://ache.pp.ru/