[ On Sunday, June 13, 1999 at 17:58:19 (-0400), John F. Woods wrote: ]
> Subject: Re: signal(SIGSEGV, SIG_IGN) -> 100% CPU 
>
> On the other hand, it always seems more troublesome to add new error returns
> to system calls down the road than to have error returns which never happen

In this case there'd be no "addition" -- it's already documented!  ;-)

> (mostly because the kind of application programmer who insists upon writing
> code to handle each and every documented error return and codes a panic for
> undocumented cases are exactly the kind of people who will gripe loudly when
> future requirements cause system calls to become more communicative about
> error returns...).

Although I can't say I've read a lot of other people's "application"
code, I would guess that the kind of error you're worried about is very
rare in most "systems" and "utility" code.  The only kinds of "new"
errno values that would cause any kind of problem would be the kind that
represent non-fatal errors, such as EAGAIN.  I have seen this particular
value cause code to fail where it shouldn't have, or never used to....

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>