On Mon, Feb 7, 2011 at 12:18 PM, Eric Hodel <drbr...@segment7.net> wrote: >> If you ignore this too, I'm going to start sending notifications to >> the list again. > > This is inappropriate, silence does not equate to success.
CI is running for the major linux platforms with failure notifications going to this list. It's done. > If there's going to be this level of argument over the CI I don't think it's > prudent to continue to use it. I'm not arguing, I made it happen. During the last developers meeting, CI was one request that came out of it. I made it happen for the three currently supported interpreters (1.8.7/1.9.1/1.9.2) on the most popular production platform (linux). The only known drawback with this environment is Hoe's incompatibility with sudo, which has an unapplied fix, and which I've pointed out multiple times and you/Ryan have ignored multiple times. This is not just a problem for CI also a problem for anyone who wants to run the rubygems build under RVM. Repeatedly pointing out that this is still unfixed is not arguing. Continuing to not just fixing the problem in Hoe IS increasing the noise on this list by forcing me to repeatedly ask for it, and you have incorrectly equated this noise to arguing. > When I make releases I don't bother to consult it, instead I run `rake > multi`. All RubyGems committers != You and Ryan > How many people have subscribed to the CI? Everyone on this list now. > Going back to the beginning of December myself, Ryan and Luis have been the > top committers. I don't see your point. All RubyGems committers != You and Ryan > If anything, our least-covered platform is Windows as Luis is our de-facto > CI. Running CI on linux isn't going to help us much. Correct, it isn't going to help much with windows coverage. It will help tremendously with immediately identifying platform-independent regressions (which are the majority, I would guess) where the committer is not you or Ryan and does not run 'rake multi' before checking in. Do not confuse lack of windows CI coverage with lack of value in Linux CI coverage, that is a non sequitur logical fallacy. Bottom line: * The RubyGems committers said they wanted CI. I have done it on linux, and will maintain it. * It is stable as of now, with failure notifications going to this list, which can easily be disabled by any committer if they go haywire and I'm not around to fix it. * A side benefit is that the build is standardized and locally reproducible via the ci_build script included in the source. Thanks, -- Chad _______________________________________________ Rubygems-developers mailing list http://rubyforge.org/projects/rubygems Rubygems-developers@rubyforge.org http://rubyforge.org/mailman/listinfo/rubygems-developers
