Ian Zimmerman wrote:
David> Ian, Since this is the first release of Poly/ML to use shared David> libraries I would prefer to give this some thought now and then David> adopt a scheme that can be used for future releases rather than David> try and change it in the future. I don't have any strong David> feelings either way and you obviously have some experience. David> Would it be best to use -version-info 0:0:0 for this release?
David> Because of the way the interface between compiled ML code and the David> Poly/ML run-time library works I'll have to think about how David> future releases should increment the version-info.
The question is if it makes sense to upgrade the runtime and standalone executables produced by linking with it independently. Not only for end users, but also for distributions.
If it does (and I think it does), then -version-info will allow that and is preferable. The rules for incrementing the numbers are spelled out in
http://www.gnu.org/software/libtool/manual.html#Updating-version-info
I thought about this and I've now made this change and updated CVS. The initial version is now 0:0 which is what the libtool manual suggests as the starting value. I hope that makes things simpler for Debian.
David.