On Jan 27, 2008 10:25 PM, Alexey Verkhovsky <alexey.verkhovsky / gmail.com> wrote: > On Jan 27, 2008 4:59 PM, Sam Ruby <rubys / intertwingly.net> wrote: > > So... the trick is to only include packages for which there are > > identified people with the interest and bandwidth to keep on top of the > > issues found. > > Exactly. In other words, committers should say "yes, we want a > continuous build, we want to be annoyed by failed build notices in our > mailboxes, and we will fix it when it breaks". Otherwise, you end up > with a permanently broken build which costs quite a bit of effort to > run, and adds little value to the project. > > If there is a commitment like that, I would volunteer to run that > build in a heartbeat. > I was running something similar for 1.8 and 1.9 build of Ruby internally at your office. A few weeks back just turned it off since was annoyed by the constant warns of FAILED BUILD warning we got. Don't want to mention that was on a Linux box, the Windows build always had 3 to 20 errors or failures all the time, that should be considered safe (or good) even with them. The complexity and the time it takes run a complete build/test procedure as pre-commit hook can make the life of the commiters a bit painful, but i think should be standard rule. Just my comments. -- Luis Lavena Multimedia systems - A common mistake that people make when trying to design something completely foolproof is to underestimate the ingenuity of complete fools. Douglas Adams