On 6/3/06, David Reiser <[EMAIL PROTECTED]> wrote: > I tried compiling g-wrap 1.9.6 outside of fink, and it worked. I > modified the 1.3.4 .info file to use with 1.9.6, and that worked, too > -- but I'm sure I had errors in the splitoff setup. So I've gone > through the packaging manual a bit, and I have some questions: > > 1.9.6 supplies libffi if configure can't find ffi.h. Do I use a > Provides: field for that, or just treat it as an internal library (no > mention of libffi in fink, possibly because the redhat pages indicate > that it is maintained in gcc now...)? >
You can probably treat it as internal then. > Many, but not all, of the packages in the 10.4 tree have revisions > starting at 1000. Since g-wrap 1.9.6 has not been available in Fink > before, do I start at 1 or 1000? > Note that the extant version has revision 1014. So start at 1000 (or 1001). Then if somebody wants it in 10.4-transitional a revision of 101 can be used, and a revision of 1 for 10.3. The reason for the separation is to force rebuilds of library packages that are built with C++ on upgrades, so that the symbols in them match what is provided on the upgraded system--thus packages that depend on the libraries can be built without getting confused about different C++ symbols in dependent libraries and on the system. > Do .a files appear in the %N-shlibs Files: field? Looking at the extant g-wrap.info as a template I'd say no. They go in the -dev splitoff instead. > > Is it true that by virtue of .so files being modules instead of > libraries, that .so files should not appear in the Shlibs: field? I believe that's the case. > Seems hard to do anyway, since there is no compatibility version > information associated with otools -L on a .so. > > I'm working in the 10.4 tree, and don't know much about the > relationships between the trees from a packaging standpoint. Should I > go ahead and submit my .info file for the 10.4 tree? > > Thanks. > -- > David Reiser > [EMAIL PROTECTED] > > > Submit it for 10.4, and get people to try it out for the other trees (10.3 and 10.4-transitional) to see if any changes have to be made. -- Alexander K. Hansen Fink Documenter (still) _______________________________________________ Fink-devel mailing list Fink-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/fink-devel
