cgal libraries have wrong install_name

[ryandesign@…]

On my Snow Leopard system:

$ otool -L /opt/local/lib/libCGAL.dylib
/opt/local/lib/libCGAL.dylib:
	libCGAL.4.dylib (compatibility version 4.0.0, current version 4.0.0)
[snip]

That should be:

$ otool -L /opt/local/lib/libCGAL.dylib
/opt/local/lib/libCGAL.dylib:
	/opt/local/lib/libCGAL.4.dylib (compatibility version 4.0.0, current version 4.0.0)
[snip]

This was  reported on the mailing list.

[raphael@…]

I have been working on a CGAL port for quite some time. My current (local) port fixes the problem of this ticket by using the cmake argument -DCMAKE_INSTALL_NAME_DIR=${prefix}/lib. Maybe this argument should be added to the cmake portgroup.

The only problem my CGAL port still has is that all Qt3 demos crash. This is the reason why I did not submit my CGAL port to MacPorts yet. A few weeks ago, I discovered that Vince already submitted a CGAL port to MacPorts. Vince, I would propose that we should merge both ports to get the best of both ports into the official MacPorts tree.

My CGAL port is attached.

[raphael@…]

Cc Me!

[ryandesign@…]

Replying to raphael@…:

My current (local) port fixes the problem of this ticket by using the cmake argument -DCMAKE_INSTALL_NAME_DIR=${prefix}/lib.

Thanks for this suggestion. I committed that fix to the cgal portfile in r63237.

Maybe this argument should be added to the cmake portgroup.

I have not evaluated all the ports that use the cmake portgroup to see if they suffer from the same problem and whether this argument would help them. But I assume they do and that it would. In that case, this is a good suggestion, but belongs in a new ticket.

The only problem my CGAL port still has is that all Qt3 demos crash. This is the reason why I did not submit my CGAL port to MacPorts yet. A few weeks ago, I discovered that Vince already submitted a CGAL port to MacPorts. Vince, I would propose that we should merge both ports to get the best of both ports into the official MacPorts tree.

That's probably a good idea, but it's hard to evaluate your portfile since I can't just run a diff against the current port. Your portfile does many things differently from the cgal portfile we already have from Vince. It might be easier to incorporate your changes if you break things out into individual patches that make individual functional changes (for example, a patch to switch to the cmake portgroup, a patch to install the additional documentation, and so on) and file separate tickets for each. That way they can be evaluated more easily and dealt with on a case-by-case basis.

I'll leave this ticket open for awhile to let Vince decide how he wants to handle your portfile submission, but I'd like to close this ticket since the issue that was reported is now fixed via r63237.

[raphael@…]

Replying to ryandesign@…:

I committed that fix to the cgal portfile in r63237.

Thanks, but I think you forgot to delete the now unnecessary post-build phase that is still in the portfile.

I have not evaluated all the ports that use the cmake portgroup to see if they suffer from the same problem and whether this argument would help them. But I assume they do and that it would. In that case, this is a good suggestion, but belongs in a new ticket.

I'll file a new ticket.

That's probably a good idea, but it's hard to evaluate your portfile since I can't just run a diff against the current port. Your portfile does many things differently from the cgal portfile we already have from Vince. It might be easier to incorporate your changes if you break things out into individual patches that make individual functional changes (for example, a patch to switch to the cmake portgroup, a patch to install the additional documentation, and so on) and file separate tickets for each. That way they can be evaluated more easily and dealt with on a case-by-case basis.

Yes, I'll file separate tickets for all outstanding issues.

[ryandesign@…]

Replying to raphael@…:

Replying to ryandesign@…:

I committed that fix to the cgal portfile in r63237.

Thanks, but I think you forgot to delete the now unnecessary post-build phase that is still in the portfile.

Ah yes. Removed that in r63340.