On Thu, Dec 2, 2010 at 9:47 AM, Philip Jackson <[email protected]> wrote: > At Wed, 01 Dec 2010 13:39:15 -0500, > Dave Abrahams wrote: > >> As *huge* fans of magit, me and my colleagues are very concerned >> about some of the recent changes. We're seeing lots of important >> functionality broken (log all, snapshot, commit with amend...) and > > Don't forget you're pulling from a bleeding-edge development repo.
Good point. Is there a stable release tag we should be using to judge the state of things? > I don't remember commit with amend being broken. It's broken from a design point-of-view, because there's no way to recover the old log message (unless it happens to be in the log message ringbuffer). > The --all bug (https://github.com/philjackson/magit/issues/issue/88) I > pushed the fix literally minutes after you reported it. I appreciate that; really, I do. > Snapshot is/was broken? * It doesn't save the index by default anymore. That's an alarming change. * Furthermore, applying a stash doesn't restore the index even if it was saved. * Just having to hit more than one key is broken from a usability point-of-view. The whole idea of snapshot is that without interrupting your workflow, you create a checkpoint that takes you back to exactly where you were no matter what else you do next. We were giving a training on Magit yesterday and discovered the problems then, which was embarrassing because we'd been raving about how wonderful and game-changing Magit is. We were even about 80 revisions back from HEAD because when I tried the HEAD the night before we immediately found log --all to be broken. When we found the snapshot problems we went back and checked HEAD but they're still there. >> recent design decisions have been pushing many highly-useful >> (formerly) single-keystroke commands into two-level pop-up buffers... > > We were simply running out of keys, really we had to do something to > free some up. Well, snapshot used to be on the `Z' key, which is now unbound. Logging has a billion options now, which is really nice from a flexibility point-of-view, but from a usability point-of-view it would have been much better (in our opinion) to keep the `l' binding for the usual shortlog case people do tens of times a day and require a prefix to pop up the grid of options for more unusual requirements. Personally I use the `log --all' option most and shortlog a little less, but I gather that pattern is not so common. > There was discussions, proof of concepts (which Óscar > kindly did) and branches which I asked people to try out and comment > on. Sorry; didn't know about this list until yesterday. Was mostly just an extremely happy user until then. > Now isn't a great time to complain about them to be honest, having > said that, there's still room for change if you can reason them well > enough. I'm tryin', here. > The other problem is there's no test suite. There was talk on > emacs-devel about including a framework officially (ert, I think) so > we can write tests when that happens hopefully mitigating regressions. Lack of automated testing seems to be a problem that affects much more than just magit. >> So I guess we're just wondering what's happening, and if there's >> anything we can do to influence the general direction. > > Code and discussion are two ways to influence development. I can see > you've made several feature requests, that's fine but if you are > willing to take the time to implement stuff you should as I'm far more > likely to just merge it. We have and will continue to submit patches. >> We've already made some contributions to magit and would be happy >> to continue to do so, but don't want to waste our time trying to >> counteract "bad commits." > > Aren't "bad commits" called "bugs" elsewhere? Well, I think I meant "regressions." > Why is fixing them a > waste of your time? Is it a waste of mine? It's one thing to fix newly-introduced bugs, but it's something else to be trying to fight against the current. I guess we just want to make sure we're not swimming upstream. -- Dave Abrahams BoostPro Computing http://www.boostpro.com
