> 	The code (as I posted the patch) definately calls fclose()
> will NULL.  The code immediately before it (mail me for a copy)
> proteects against the NULL case for some other f...() calls... which
> leads me to believe that fclose(NULL) is accepted by some OSs.

fclose(NULL) is accepted by some OSs, but NetBSD made a conscious choice
to not accept it.  It is not correct, and I believe that the prevailing
opinion was that not accepting NULL would help uncover incorrect programs.

Personally, I'd probably have allowed it--or perhaps used assert(), but
for that to be effective, we'd need a debugging version of libc...

-allen

-- 
              Allen Briggs - end killing - briggs@macbsd.com