On 04/04/2013 04:51 PM, Lucas Meneghel Rodrigues wrote: > (this was also sent by me using my work account, but today doesn't look like a > great day for the company's smtp servers, so I've decided to send it twice). > > Hi folks, > > As we're always looking to streamline autotest and the offspring projects > (such as virt-test) development, one thing that occurred > to us is that code reviewing/patch management done through the mailing list > has disadvantages [1]: > > 1) Downloading/applying/reviewing patches is not very convenient. We used to > have a public patchwork instance that had to be disabled as we decomissioned > test.kernel.org <http://test.kernel.org>. We have an internal patchwork > instance, but as the name says, it is internal, therefore we alienate people > that are not from Red Hat to conveniently download and apply patches. > > 2) Reviewing means a long thread of emails, with no visual cues to help the > reviewer, such as colors. > > 3) Following patch series is harder (you keep making searches on your mailing > list folders to look for previous versions of a patch and all that). > > 4) We already use github pull requests, so having to scan for new work items > on the ML and github is kind of an overhead. > > I'd personally like to have something more gerrit > (http://code.google.com/p/gerrit/) like on github, but the pull requests > infrastructure is 'good enough' for what we want, and we don't get to maintain > infrastructure, which is a big "yay!" for us. > > Therefore, I'd like to move to a model where we mandate pull requests for > contribution to autotest, virt tests and projects inside the autotest > umbrella. Of course, I don't want you guys to think I'm shovelling this down > your throats, so I'd like to hear if anybody feels very strongly about it. > Also, we could consider exceptions in cases like: > > * I hate github so much that I can't stand using it and I think you are all > morons for even considering this OMG. > * I can't use github because, I don't know, my company forbids using web > browsers or send any requests to the port 80 or 443, or whatever. > > So please, let us know what you think. By moving to a single, concentrated > point of review I believe we'll have less development overhead and generally > reduce stress levels of the maintainers. >
Sounds fine to me FWIW. I'd recommend for anyone looking to take a bit of the edge off using github to look at: https://github.com/defunkt/hub#readme Enables submitting pull-requests and other bits in one shot from the command line. However, I recommend _not_ using their recommendation of alias 'git' to 'hub' , instead just install it as a git subcommand (copy it to /usr/libexec/git-core/git-hub on Fedora at least). With their recommended configuration the alias causes massive slowdown if using git-prompt.sh in big repos. - Cole _______________________________________________ Virt-test-devel mailing list [email protected] https://www.redhat.com/mailman/listinfo/virt-test-devel
