Re: AC_FUNC_SETVBUF_REVERSED in cross-compilation

[Paul Eggert]

> From: "Paul Smith"<address@hidden>
> Date: 09 May 2002 22:12:17 -0400
> 
> Just curious, but what does this do on systems which don't have
> setvbuf() at all?

It doesn't define the symbol.  Not that I have any systems to test
this on.  It's been quite some time since I used a system that lacked
a standard setvbuf.  I think the last one was probably produced about
10 years ago.

> From: "Paul D. Smith" <address@hidden>
> Date: 09 May 2002 22:16:47 -0400
>
>   eb> Well, I was about to ask for a list of which systems are broken
>   eb> and suggest that a better solution would be to make the value
>   eb> hardcoded based on a filter of that list rather than trying to
>   eb> determine it via executable testing.
> 
> Hm, not a bad idea.

I think it's a bad idea.  There's no good way to maintain the filter.
It used to be done that way, 10 years ago maybe; people used "#ifdef
Xenix" and junk like that.  But even when the issue was fresh in
people's minds, they still screwed it up.  (How many people remember
that SCO Xenix 386 2.2 reversed the setvbuf args, but 2.3 did not?
Or what about the AT&T 3b1?)

Hence the automated test; it's far more reliable.

> The comments say this was a problem on "System V before release 3".
> Unfortunately I have no idea how to test that based on the system
> type.

Yes; that's why it's a bad idea.