[PATCH] *BSD patches plus dbus-launch fix for regression tests

[John (J5) Palmieri]

On Tue, 2006-10-24 at 20:59 -0400, Havoc Pennington wrote:
> Hi,
> 
> Does this add -lpthread in the .pc file on Linux? That's something I 
> believe we're supposed to avoid, though hell if I understand what the 
> deal with libpthread is.

On Linux it doesn't.  The deal is some functions are in glibc (the
internal mutex and condvar stuff) while the thread creation method
(client methods) are still in lpthread on linux.  On *BSD the mutex and
condvar funtions are in -lpthread.

> 
> The dbus-launch change is only partly DBUS_BUILD_TESTS conditionalized. 
> I think a less intrusive patch would use an env variable, then you can 
> skip all the arg parsing etc. - in fact there may already be some 
> relevant env variable like the build dir set in a test context? What if 
> you just export the DBUS_TOP_BUILDDIR or something if isn't, from the sh 
> script calling dbus-launch?

Could do but that means we will be executing a path we know will fail
every time the user calls dbus-launch, but I guess we should be
compiling with tests off for the final build anyway.  

> You could also maybe just fall back to execlp #ifdef DBUS_BUILD_TESTS, 
> something I think we do elsewhere, then set the path in the script.
> 
> The HAVE_CMSGCRED change isn't everything that I meant; what I was 
> suggesting is that we add a HAVE_LOCAL_CREDS, and adding a configure 
> check to define that. And do configure in such a way that only one of 
> the two will ever be defined. The patch you have looks good fixing the 
> existing code, but the configure change would keep us from doing it 
> wrong in the future.

I thought of that after I sent out the patches but was too lazy to fix
it up right away ;-).  I'll do that tomorrow.

> Thanks for subjecting yourself to freebsd for a noble cause ;-)

It wasn't so bad once I found a LiveCD that would boot.  Useful little
suckers.  These little inconsistencies are a pain in the butt and I wish
we had a more vibrant *BSD community.  The fixes were pretty trivial
once one could actually sit down and reproduce the issues.

-- 
John (J5) Palmieri <johnp at redhat.com>