>> >No, the one in /usr/lib is a link to the one in /lib, but they are >> >not the same files, but that isn't relevant. But why is there no >> >libdb-3.so file (if we are going to go with the naming used for >> >libdb40).
>> >>What's the problem exactly? >> >The fact that the versioned library file doesn't have it's version >> >trailing the .so. >> Yes. > So what should happen if a package has a build requirement on a specific > version of a library (e.g. libfoo.so.1.0.7)? Then you'll need to specify this specific BuildRequires in the .spec file, with the package name. BuildRequires: libfoo-deve >=1.0.7 or something like that. > Or is this impossible unless a versioning scheme like with db3 & db4 is > used? If it should work you're right to want different dependency names. We're seeing that more and more of these dependencies have version numbers in them for the major part. Good example are the glib1.2-devel and glib2-devel packages. >> >Stefan, I think the problem isn't the naming of the requirements, >> Well, you can't disagree that using the ".so" is at least confusing. >> It's very difficult to tell where it came from. >> If we would change the naming, then even _if_ packaging is incorrect, >> it wouldn't fool the system --> and that's one of my main motivations >> to change the naming. >> >it's the method. We shouldn't be chopping off the bit that follows >> > .so, we should be finding the file that links to the library listed. >> I don't follow you here... > For a regular dependendy to libdb-4.0.so one might expect a -devel > dependency on libdb-4.so, or even libdb.so. Note that libtool doesn't > get this right either (or does it use the same code?), judging from > /usr/lib/libdb-4.0.la: > library_names='libdb-4.0.so libdb-4.0.so libdb-4.0.so' > > At least, I would expect it to be: > library_names='libdb-4.0.so libdb-4.so libdb.so' Yes, this is confusing. And I don't intend to these inconsitancies :-) As unix is consistanly inconsistant. I'm aiming to make the -devel dependencies smart enough that it won't be fooled by these inconsitancies. By introducing a different naming scheme the confusion is over... Stefan
