Jonathan Adams wrote:
>> So would it be reasonable to assume that so's that don't have the
>> version suffix might be dlopen candidates? Maybe I'll just exclude them
>> from the removal algorithm?
I don't think you can infer much from a name.
Our founding fathers had a design. Shared objects would have major
numbers (ie. libfoo.so.1), and symlinks would be established to these
(ie. libfoo.so -> libfoo.so.1), so the developers could reference
these dependencies at compiler time using the -l notation (ie. -lfoo).
A shared object that can be specified as a dependency (as created when
using -l), is no different from a shared object that can be dlopen'ed.
Some applications might have dependencies on libfoo.so.1, others might
wish to dlopen("libfoo.so.1").
The conventions have gotten very lax. We have shared objects that
are only delivered as "*.so". Shared objects that are only delivered
as "*.so.?". Both. Or even various incantations of "*.so.1.23.65.99".
I fear that whatever naming algorithm you use, you'll leave some
objects behind that could be deleted, and delete some that are still
required :-(
--
Rod
