> > As part of my goal of making log4cxx ABI compatible, I've added a new check > > to the Github actions that will run automatically to see if there are any > > ABI changes. Currently it's just sitting in a PR( > > https://github.com/apache/logging-log4cxx/pull/58) if anybody wants to > > check it out before I merge it.[...] > > Just to be sure I understand correctly: There's always an "abi.dump" > in the repo as how the ABI should look like. For each new commit the > GitHub action generates "new-abi.dump" to compare against and fails in > case of differences. If the ABI should be changed by purpose, the > action can be made aware by simply committing a newer version of > "abi.dump" as part of a change/PR. > > Correct? >
Yes, that is correct. > > On a somewhat related note, since we are now making use of the new C++11 > > features(shared_ptr, mutex, etc), what should the goal of our next release > > be and when would it make sense? I ask because we could go for the next > > release being ABI-compatible, although it would be rather tedious to do > > that - there's a lot of code that would need to be changed, but it's not > > particularly hard. > > This reads to me like if you want to release pretty soon what is > available now? If so, I totally agree, otherwise I would be fine with > as well. ;-) > > Everything you have changed right now should make the lib more stable. > Just think of the memory leaks and thread-issues which should be fixed > now. OTOH, it has the risk of breaking things for users with older > toolchains like me. THOUGH, even then it's good to make people aware > of that early with a release right now. So make a 0.12.0 now, so that > fixes mostly related to C++11 etc. can be published as 0.12.X and keep > ABI-compatibility for 0.13 or alike. > That sounds like a good idea to me. I'll start doing that in a week or two, I think there's still some documentation that I want to make sure is updated. That should also give it a period of time to allow for any new bug(s) to be reported. I have some applications on my side as well that I'll compile with and make sure that they still work correctly. -Robert Middleton