Junichi Uekawa <[EMAIL PROTECTED]> writes: >> This approach would make the assumption that any lib*-dev package >> required to build a package containing a library should also be a >> dependency of the development version of the library. Putting aside >> whether it's worth keeping such a dependency only for static linking >> when the one library completely hides the usage of the other, this >> fails for large packages that produce multiple binary packages. >> Consider a package that builds a library that depends on nothing but >> libc and a separate binary package of executables that depend on both >> it and, say, LDAP libraries. The LDAP -dev package will be in >> Build-Depends, but the library -dev package should not depend on it.
> Such case is an exception, which needs to be properly done, and is not > something which is usually done. As you have explained yourself, the > exceptional case is: > 1. when there is no static library Well, this problem arises for any source package that builds multiple binary packages where the build dependencies for the package itself aren't needed for the library, not just for those where there is no static library. I only maintain a small number of libraries myself and yet several of the packages I maintain would be affected by false positives here. > and > 2. when the library hides the shared library it uses from the end-users > of the library. This is also more common than not. My guess is that you'd get a false positive rate approaching 10% for this particular test, unfortunately. That seems too high to me for it to be worth it, although I admit this is a problem that would be hard to identify via other means. -- Russ Allbery ([EMAIL PROTECTED]) <http://www.eyrie.org/~eagle/> -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
