On Thu, Jan 09, 2014 at 11:09:09AM -0500, Stephen Gallagher wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 01/09/2014 09:39 AM, Jan Engelhardt wrote:
On Thursday 2014-01-09 14:42, Simo Sorce wrote:
Sorry, but in my book changing the SONAME for a backwards compatible change is not ok [...] I do, maintaining a stable SONAME is important [...] The only solution for backwards compatible changes that require a new symbol with the same soname is to have an explicit version dependency in the package management system. That can be enforced only at build time.
Hey, keeping SONAME and doing symbol version maps was my preferential pick as well, but Lukas Slebodnik did not want to. Come on, is it really _so much_ work to maintain a symbol version file? How many symbols get added anyway?
The conversation got sidetracked by the soname bumping discussion (which is a terrible idea and should be dropped immediately). Imagine if glibc bumped soname every time it added a function...
I agree that maintaining the table shouldn't be a huge added effort, but one that we do need to codify into our release process documentation so we don't ever miss it.
Just give me a clear answer already. Will sssd address the issue, or is it going to be a "fuck you" to downstream packagers?
That kind of language is unnecessary and offensive. Please tone it down.
We have been discussing the merits of symbol versioning today in the IRC channel. Given the advantages inherent in being able to auto-detect these versioned dependencies in both the RPM and DEB build-systems, adding this version map seems to be the best approach.
Thanks for having the discussion on IRC! Would you (or Lukas, or Simo) like to file a ticket for both SSSD and ding-libs ?