On Sat, 04 Nov 2017 22:23:32 -0400, Matt Harbison wrote: > On Sat, 04 Nov 2017 22:11:21 -0400, Yuya Nishihara <y...@tcha.org> wrote: > > My concern is that the doctest is too limited in scope to cover possible > > bugs > > caused by Windows drive letter. Maybe we could add a safer alternative to > > os.path.relpath() and ban the use by check-code. > > Seems like a good idea. There are a few more uses of it, and I'm pretty > sure that at least cmdutil._conflictsmsg() is broke too, based on passing > repo.root and cwd. (I guess thg doesn't use this function?) > > I haven't been able to create the issue with the origpath code, and like I > said in the 'share --relative' patch, the error is useful there. Is this > one of those things we can ban, but just record the few places it's needed > in test-check-code.t?
Perhaps os.path.relpath() can be replaced with two functions, which are: a) always return a relative path; raise Abort if not possible b) try to return a relative path only if possible; an absolute path otherwise On Unix, they both are mapped to os.path.relpath. _______________________________________________ Mercurial-devel mailing list Mercurial-devel@mercurial-scm.org https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel