Re: Pthread detection in glib 2.2.1

[Owen Taylor <otaylor redhat com>]

On Mon, 2003-03-10 at 15:38, Daniel Egger wrote:
> Am Mon, 2003-03-10 um 20.32 schrieb Owen Taylor:
> 
> > Anyways, I spent a ton of work on cross-compiling prior to
> > 2.2.0, so if you haven't tried cross-compiling since then,
> > I'd advise trying it and/or reading the source code.
> 
> Actually I have both and have to say that I'm generally 
> very pleased by the work you've done including applying
> other work in this area. Thanks.
> 
> My intention here is to make sure that dubious problems are 
> avoided, asking the linker to check for a function is 
> certainly better than running testcode (or failing in case
> of crosscompilation) by the means that the existance of the
> exercised function is checked. However there are cases where
> checking for existance of a function is *not* enough to also
> grasp for correct semantics and threading implementations are
> a very nice example for this. Better would be to check with
> the compiler/assembler/linker for known features/semantics
> because they certainly know best (of the tools we have access
> to in the moment of configuration).

I don't quite get how you can check a feature of the _runtime_
environment when cross-compiling. 

 uname -a | grep 'Brokenix.*4.2'

isn't going to work any better than an AC_TRY_RUN. 
Plus, when you find a buggy version of an operating system, you
frequently have no idea what the right range of version
numbers is for such a check.

Having the user set cache variables manually works as a 
technique of last resort, but there is no evidence that that
is necessary here.

Regards,
                                    Owen