readline-5.0 compiles on OSX Tiger after applying a small patch to get rid of a "powerpc-apple-darwin8-gcc-4.0.0: -compatibility_version only allowed with -dynamiclib" message Reproducible: Always Steps to Reproduce: 1. cd /usr/portage/sys-libs/readline 2. ebuild readline-5.0-r2.ebuild compile 3. Actual Results: compilation died with the message: powerpc-apple-darwin8-gcc-4.0.0: -compatibility_version only allowed with -dynamiclib Expected Results: finish compiling without errors supplied patches allow compilation to go fine and the files to actually be usable (32-bits only)
Created attachment 62089 [details, diff] readline-5.0-r2 ebuild patch patch which enables osx patch and elibtoolize
Created attachment 62092 [details, diff] patch which enables readline to compile on OS X located in files, this patch fixes a shared object config file to include -dynamiclib where -compatibility_version is used
readline-5.0-r2.ebuild has no USE flags, nor does it use any global USE flags.
Is this a -collision-protect package? I was under the impression that OSX had a gimped version of readline already. If this is a -collision-protect package, did you just build it in /var/tmp/portage, or did you actually overwrite Apple's stuff and test it? (Note - no one requires you to do that, I just need to know for keywording purposes).
yes. This package installs two files that collide with Apples files. I thought I needed readline, but probably could have inserted a system.provided line as well. Anywayz, I indeed turned off collision protect for this package. Life would get unpleasant but not unbareable if it would break... my risk here.
Comment on attachment 62092 [details, diff] patch which enables readline to compile on OS X patch is wrong
Created attachment 67005 [details, diff] readline-5.0-r2.ebuild.patch Skips all the silly /lib and ldscript hackery for userland_Darwin. Not sure if you want to use the use-expanded var as a conditional or not...
i guess the building issue was resolved with Bug 99143 ? if so, your proposed diff looks good, feel free to apply it
kito's patch seems not to address the ppc-macos build problem, which my attachment #62092 [details, diff] addressed: gcc version 4.0.0 (Apple Computer, Inc. build 5026) powerpc-apple-darwin8-gcc-4.0.0: -compatibility_version only allowed with -dynamiclib make[2]: *** [libreadline.5.0.dylib] Error 1 make[1]: *** [supported] Error 2 make: [shared] Error 2 (ignored) (note, although gcc fails, the ebuild just continues, because make doesn't fail)
Ok, should be all fixed up inCVS. Fabians errors were a result of bad CHOST being set.