Stefan Kangas writes: > Philip Kaludercic writes: > >>> If we really can't find a better option, we need to add a bunch of >>> comments explaining why we ended up with such a hack. >> >> It seems that it is only necessary to bind `package-alist' during the >> invocation of `package--used-elsewhere-p', so this is a more >> conservative proposal: > > Any chance we could add a unit test for this? It's been a while since I > last looked at package-tests.el. > > Other than that, LGTM. How does this look like: