On Mon, 2016-02-29 at 17:21 -0500, Jeff King wrote: > On Mon, Feb 29, 2016 at 05:10:24PM -0500, Jeff King wrote: > > > > We apparently don't always call check_repo_format before calling > > > git_config_early -- or, more to the point, before doing ref > > > operations. > > > So I think we need this in git_config_early. > > > > That seems horribly broken, though. If a program is accessing refs > > without calling check_repository_format, isn't it bypassing all of > > our > > regular version and extension checks? > > > > I think we should be smoking out such cases (e.g., by setting > > a null refs-backend, as I mentioned earlier) and fixing them, > > rather > > than working around them by putting the backend setup in the wrong > > place. > > So even without your series, this isn't too bad to instrument, like > the > patch below. > > It does reveal a handful of failures in the test suite. I haven't dug > yet, but I strongly suspect those are all bugs that should be fixed. I'll try to add a patch for these cases. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html