Hi, I have seen various packages (mostly from the same maintainer, though) which do branch updates in a imho wrong way.
Updates to a stable branch fixes or backporting fixes is OK. I don't deny that or so. But the package IMHO should have a correct version then. e.g. if you have a package 1.0 and add a complete branch update as a patch (or upgrade to a snapshot) one should do a 1.0+gitYYYDDMM-1 or whatever format you choose. Not 1.0-15 or so. That's what I just saw on debian-devel-changes: On Sat, May 28, 2016 at 04:48:47PM +0000, Matthias Klose wrote: [...] > Changes: > gcc-5 (5.3.1-21) unstable; urgency=medium > . > * GCC 5.4.0 release candidate 1. > * Update to SVN 20160528 (r236840, 5.3.1) from the gcc-5-branch. > - Fix PR libstdc++/69703, PR libstdc++/71038, PR libstdc++/71036, > PR libstdc++/71037, PR libstdc++/71005, PR libstdc++/71004, > PR libstdc++/70609, PR target/69634, PR middle-end/68142, > PR middle-end/69845, PR rtl-optimization/68814, PR lto/69003, > PR ipa/66487, PR target/69252, PR target/67973 (x86), > PR middle-end/67278, PR target/67278 (x86), PR tree-optimization/69720, > PR tree-optimization/67921, PR middle-end/70941, PR middle-end/70931, > PR tree-optimization/70623, PR tree-optimization/70780, PR c++/70347, > PR c++/70466, PR fortran/71204, PR fortran/69603, PR libffi/65567, > PR libstdc++/70762. > * Update the ibm branch to 20160526. A 5.4.0 rc1 in a package versioned 5.3.1-21? This is even worse since the last gcc/python/... upgrades which "just" upgraded to SVN revision X without bumping the base version. I'll be attacked nevertheless claiming this would be attack but this is NOT against a specific maintainer though somewhow that mostly happened in those packages, but a general problem. Package versions should actually tell the correct version... I think we should amend the policy to make that clear. Regards, Rene
