On Wed, Jan 14, 2009 at 1:24 PM, Eric Wilhelm <[email protected]>wrote:
> And, anything that does turn-up at the pre-release stage is probably > several commits removed from the change which caused it, so the > committer is either MIA at that point or has lost significant context > compared to within several minutes or even a day of the commit. This > leaves me diagnosing the problem, running svn blame, pulling the log, > sorting intent from bug, and in the many cases writing unanswered > e-mail and thus losing my own context on the matter. This is very > frustrating and inefficient. > I greatly appreciate that you've taken on this. But please don't think you've got to do it alone. If we get enough people willing and able to patch things, then we can spread the work around some. > And the test coverage is only about 18% (if you ignore cross-platform > issues), so this sort of trouble is all too likely. > That's a bigger and broader problem. > Send an error report immediately after the commit, and the committer > (who may know exactly what went wrong without looking) will probably > still be available to fix it. > What if we moved to a more regular alpha release schedule? That would get CPAN Testers involved to check things. That's not quite granular to each commit, but it will speed up the feedback cycle quite a bit. -- David
