Re: is --enable-threads supported on HP-UX 11.00?

[Mark Mitchell <mark at codesourcery dot com>]

> This could be the case if the OS actually supported some thread
> library, but GCC didn't.  I don't know of any OS for which this is the
> case.

Really?  What about cases where the only thread support is non-POSIX
weirdness?

But anyhow that is not somehow an invariant; that is only because
we've remembered to implement the threading support.  If a platform
comes along where we have not, then we will be in this situation.  In
other words, we are not programming defensively if we just assume
that we always implement thread support.

> OS doesn't support any thread library at all.  In this case,
> --enable-threads is totally pointless, and may be safely ignored,
> IMO.

I think, again, that this is some key difference in perspective.  I
can't really say that you're wrong -- it's clearly a value judgement.
It's just that this seems analagous to writing:

  extern "C" { auto int i; }

You could say "well, `auto' doesn't make any sense there, so I bet
you didn't mean that, so let's just ignore it, but not issue an
error."  I'd rather you said "hey, bozo, extern variables can't
be auto."  If I'm dumb enough to configure with --enable-threads
on a platform with no threads, I'd like to know about it; I might
be planning on trying to *use* the non-existant threads!

(Far worse than this is that configure accepts, and ignores,
--enable-theads, --enable-threadss, and such.  I know why it's hard
to get this right, but it's still rather pathetic from a usability
standpoint.)

I seem to have been on a lot of crusades lately.  How obnoxious.
I think I need to focus on the crusades that matter most; I'll
call the knights home on this one, if you still don't agree.

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com