Bill Allombert wrote: > On Tue, Apr 06, 2010 at 10:35:58PM +0300, Yavor Doganov wrote: > > I think that *provided* the adaptors are built from the same > > source package, it is OK to trim their dependency on > > libgnustep-dl2-0, as it > > Being from the same source package does not make any difference for > binary packages when upgrade are concerned.
Ah, right :-( > > is guaranteed that they will migrate together to testing and a prudent > > user (except those manually installing .debs) will be guaranteed to > > have compatible adapror + libEOAccess. > > This is not OK, because partial upgrades need to be supported and I > am not sure if the current situation (before trimming) is safe with > regard to partial upgrades: > The only versionned dependency is > gnustep-dl2-sqlite-adaptor (0.12.0-1) Depends: libgnustep-dl2-0 (>= 0.12.0) > So someone can upgrade libgnustep-dl2-0 to 0.13.0 and keep > gnustep-dl2-sqlite-adaptor to 0.12.0-1. Does that work ? Yes, it will work if there are only API additions in 0.13.0. > Be careful to avoid this: > libgnustep-dl2-0: Depends: gnustep-dl2-postresql-adaptor (=0.12.0-1) | > gnustep-dl2-sqlite-adaptor (=0.12.0-1) > because this allows a system to carry gnustep-dl2-postresql-adaptor > 0.12.0-1 and gnustep-dl2-sqlite-adaptor 0.11.0-1 and > libgnustep-dl2-0 will likely attempt to load > gnustep-dl2-sqlite-adaptor 0.11.0-1. If an old adaptor does not work with the new version of the libraries, this implies an ABI break in which case the SONAME would have to be bumped. In fact it is a good (but far from complete and guaranteed to succeed) practice to test the old adaptors against new upstream releases. In any case, if the ABI break remains undetected, the library will still load at least 1 adaptor successfully and will not abort as if no capable adaptors are present. > I would suggest to implement some simple mechanisms to define the > interfaces and have libgnustep-dl2-0 only look for adaptors for the > correct interfaces. This has to be discussed with upstream... -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
