Hi Tim, On Tue, Aug 28, 2018 at 10:27:34PM +0200, Tim Düsterhus wrote: > Willy, > > Am 25.08.2018 um 08:13 schrieb Willy Tarreau: > > Done,thanks. > > > > I just noticed that the reg-test still carried the old name (h*), > instead of the new one (b*), because Frederic renamed them in the mean > time, while this patch still was pending. You should rename lua/h00001.* > to lua/b00003.*.
Well, that's exactly why I prefer patches over instructions. Having to apply various operations spread over multiple e-mails to some data is confusing and time-consuming. Often it's much easier for the reporter to simply update a patch than to explain all has to be done. And sometimes it even highlights remaining issues that are not necessarily noticed before the patch is written. > And looking at the file names I am thinking whether the current naming > scheme of an incrementing number is a good one. It will probably cause a > ton of merge conflicts (when reg-tests are regularly provided) in the > future. I thought exactly the same while merging it, but given that we're almost all convinced that the reg-test directory's organization will continue to evolve a bit until we find the best one, I was not too much concerned. > It might make sense to either use the timestamp for the > filenames or a date + short slug representing the test description: > > b_20180828_txn-get-priv-scope.vtc That could be an option indeed, especially for bugs. Another aspect I was thinking about is to backport the reg-test stuff once we're pretty satisfied with the organization so that we can use it to check if a version is affected by the bug and if the backported fix properly addresses it. We just need to remind that sometimes these tests will have to be adapted during the backports, either because the original one uses options not available in the older versions, or just because it isn't triggered exactly the same way. Cheers, Willy