REVISION: zlib universal

[ryandesign@…]

Since we seem to be on a push to make universal ports optionally available, and I've had some experience building zlib universal, I wanted to provide a patch to add a +universal variant to the zlib 1.2.3 portfile.

[ryandesign@…]

add a +universal variant

[eridius@…]

I took a look at that patch, and it's pretty ugly.

Here's one that cleans it up, outputs messages to the user to tell them that it's building libz.a after destroot, and indents everything better.

[eridius@…]

Incidentally, the original patch didn't even work (the reinplace was broken).

[ryandesign@…]

Eridius: Thank you for cleaning it up. You obviously know a lot more tcl than I do. I was hoping there would be a more elegant way than what I did.

I would however prefer to keep patches to a single purpose. That is: I'd prefer we not make whitespace changes and also functional changes at the same time. It makes it very difficult to later figure out what the substantive changes were, if svn diff shows the entire file to have changed due to differing whitespace.

FYI: My original reinplace did work, but required GNU sed, which I already had installed; I didn't realize BSD sed which Mac OS X provides doesn't support the \0 notation. To support BSD sed, I had also reached the same solution as you, grouping the entire search pattern and using \1 in the replacement.

Pipping: It looks like you only removed "-Wl,-syslibroot,/Developer/SDKs/MacOSX10.4u.sdk" from LDFLAGS? An older version of the Apple tech note recommended using that, and I seem to still have a habit of using it, though the current documentation no longer lists it, and it seems to work just fine without it in this case.

In summary: I would be happy if zlib-universal-third.diff were applied. (Well, I'd be happier if functional changes and whitespace changes were made in two separate commits.) Now we just need approval from the maintainer. Landon, what do you think?

[ryandesign@…]

Looks like Landon has abandoned the zlib port. I'll took it over in r22499.

[eridius@…]

Until somebody tests it on PPC, I would not recommend removing the -Wl,-syslibroot because the SDK is irrelevant on i386, and so I'm not willing to risk breaking cross-compilation from PPC.

That said, as you made zlib openmaintainer, I'm going ahead and committing my version. I know, it would be nice to separate whitespace changes from functional, but it's hard to do that at this point, and there's also far more functional changes than whitespace changes.

Committed in r22776.

[ryandesign@…]

I am on PPC and I did test it there.