On 2014-11-01 16:50, intrigeri wrote: >> As I recall, we generally prefer libraries do not have unnecessary >> > "strictly equal runtime version checks", since they tend to be wrong. >> > Proper use of shlibs (or symbols) and SONAME bumping (with package >> > renaming) makes such checks redundant (for Debian maintained reverse >> > dependencies). > Actually, the way I understand the code, what we have here is not > a "strictly equal runtime version check", but rather "runtime version > greater or equal to the build-time one", which is better in that it > doesn't require binNMUs.
For reference, should a check would still be wrong with symbols files. If libotr had been using symbols files, a reverse dependency could build against 4.1.0, but only have a dependency on >= 4.0.0 (provided that they did not use the "otrl_mem_differ" symbol). Combine that with the case where libotr/4.0.0 is in testing, libotr/4.1.0 is stalled in sid and the reverse dependency migrates to testing => reverse dependency is now broken in testing due to no fault of its own. It would now require binNMU in testing simply to get rebuilt with a "built-against 4.0.0" marker. ~Niels -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
