Shibboleth Developers

[Scott Cantor <>]

Lukas Haemmerle wrote:
> I think the errors are the same as the ones we saw earlier:
> The process has forked and you cannot use this CoreFoundation 
> functionality safely. You MUST exec().

I'm confused. I thought this had been tested by somebody running an Intel 
Mac. If we build Xerces with no CF calls, and it's still not working, then I 
think we're unable to support Leopard unless Apple fixes this bug.

But what code is left that's calling these APIs? Did you build Xerces using 
my portfile?

> However, this problem is more related to Leopard than to the 
> architecture problem mentioned in my last mail.

Yes, I understand that, but I thought it was solved.

> What I also found a bit strange is that Mac Ports
> apparently only compiles for the 32bit architectures ("-arch i386 -arch 
> ppc") but not the 64bit version or am I missing something here?

I think Apple is flat breaking their OS here. If you ship 64 bit code and 
expect people to build 64-bit libraries, then your toolchain (remember, the 
gcc here is Apple's) should build 64-bit by default. That's what 64-bit 
Linux does as far as I can see.

Maybe this is a libtool issue though.

> If true, these 32bit universal binaries wouldn't help much I guess...

Nope. I was far more concerned about the bitness issue. If you build on 
Intel, you should get Intel binaries and I see no real need to make them 
universal. That's why I just did source-based portfiles, I expected that 
people would build on the machine where it was being used, so CPU arch 
shouldn't be an issue.

But to fix this, we would probably have to do a platform variant that sets 
CFLAGS/CXXFLAGS to force a 64-bit build.

BTW, I believe I've fixed the version macros in the portfiles that were 
broken.

-- Scott