On 16 May 2005 at 21:25, Steve Langasek wrote: | Package: libquantlib0 | Version: 0.3.9-1 | Severity: serious | | Hi Dirk, | | Reviewing the changes in quantlib 0.3.9-2 has brought this fact to my | attention: | | $ dpkg -c q/quantlib/libquantlib0_0.2.1.cvs20020322-1_i386.deb |grep libQuantLib | -rw-r--r-- root/root 2815996 2002-03-23 14:49:03 ./usr/lib/libQuantLib.so.0.0.0 | lrwxrwxrwx root/root 0 2002-03-23 14:48:43 ./usr/lib/libQuantLib.so.0 -> libQuantLib.so.0.0.0 | | $ dpkg -c q/quantlib/libquantlib0_0.3.9-1_i386.deb |grep libQuantLib | -rw-r--r-- root/root 5192416 2005-05-02 22:22:23 ./usr/lib/libQuantLib-0.3.9.so | -rw-r--r-- root/root 208464 2005-05-02 22:22:23 ./usr/lib/libQuantLibFunctions-0.3.9.so | lrwxrwxrwx root/root 0 2005-05-02 22:21:47 ./usr/lib/libQuantLib.so -> libQuantLib-0.3.9.so | lrwxrwxrwx root/root 0 2005-05-02 22:21:57 ./usr/lib/libQuantLibFunctions.so -> libQuantLibFunctions-0.3.9.so | | The soname of the quantlib library has been changed, but the package name | has not. Since there are packages in woody which have a Depends:
That's been the case since upstream started encoding the soname in the package name some time in the 0.3.* release series, yes. | libquantlib0, this is a release-critical bug because it invalidates the | shlibs from libquantlib0 in woody and breaks partial upgrades between stable | releases. E.g., consider a user who has quantlib and quantlib-python | installed on woody, begins transitioning to sarge, and pulls in | quantlib-ruby -- the quantlib-python package will be silently broken by the | new libquantlib0 package that is pulled in. In theory, yes. In practice, only the three packages I just uploaded (quantlib-ruby, quantlib-python, rquantlib) depend on libquantlib0. | Please rename libquantlib0 so that the package name reflects the soname of | the library it contains (e.g., libquantlib-0.3.9). I really do not want to do this because the soname will change for every new upload of quantlib. This would flood the archive with meaninglessly redundant packages I would have to run after. Can we address this some other way? Dirk -- Statistics: The (futile) attempt to offer certainty about uncertainty. -- Roger Koenker, 'Dictionary of Received Ideas of Statistics' -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]