Redefining C Builtin Functions

[gnu at hoptoad.UUCP]

In article <276 at usc-oberon.UUCP>, blarson at usc-oberon.UUCP (Bob Larson) writes:
> In article <2528 at brl-smoke.ARPA> gwyn at brl.ARPA writes:
> >                                               The only
> >relatively safe redefinition would be one with a standard-
> >conforming interface, but then why not just use the one
> >provided?
> >
> Perhaps because library authors arn't always perfect, or just for
> performance.  (Is your strcmp optimised for 10Mbyte strings?)

Most of the time when I try this, it doesn't work though, due to the
"1 source file = 1 loadable module" semantics of Unix linkers.
If I try to replace one routine with a non-buggy version, I end 
up having to replace the whole source file, containing some large
number of routines that I might not even have source for.

What does the ANSI C standard say about this?
-- 
John Gilmore  {sun,ptsfa,lll-crg,ihnp4}!hoptoad!gnu   jgilmore at lll-crg.arpa