cvs commit: src/usr.sbin/sysinstall main.c

[Alfred Perlstein]

Using the strategy "commit to -current then suffer the fallout"
is pretty bogus.

I don't understand why some form of compatibility or #define wasn't
thought out before hand.

This stands out like "fixing select" to record time elapsed into
the timevals, POSIX'ly correct, but incorrect for FreeBSD, without
more thought.

* Andrey Chernov <ache at FreeBSD.org> [070430 18:07] wrote:
> On Tue, May 01, 2007 at 04:59:42AM +0400, Roman Kurakin wrote:
> >  Hi,
> > 
> >  Since there is some noise around this, could we just scream for a while
> >  that code should be fixed but allow it to still work? After some time than
> >  the majority of the buggy code will be  fixed we will stick to the std
> >  behavior? IMHO this will be less painful.
> 
> Nice idea, but depends of amount of problem reports I'll got. If the 
> number will be small, there is no pain to be std. After my last commit I 
> not get any report yet (excepting unclear "some two ports", without 
> names).
> 
> BTW, screaming can be only for args checking. For putenv() right or 
> wrong usage can't be detected inside lib.
> 
> -- 
> http://ache.pp.ru/

-- 
- Alfred Perlstein