On Fri, Sep 26, 2008 at 8:23 AM, Dmitry Kurochkin < [EMAIL PROTECTED]> wrote:
> On Fri, Sep 26, 2008 at 3:17 PM, Dmitry Kurochkin > <[EMAIL PROTECTED]> wrote: > [skip] > >> > >> Dmitry's code tells us that if we are calling ioAbsolute on a file, then > >> we must get its parent directories' ioAbsolute name. One question: what > >> happens if the file in question is "/foo"? Does super_name do what we > >> expect? > > > > Yes. I did not test it. But is should work fine - super_name returns > > '.' in this case. > > Actually, after thinking on this a bit more, it is broken for "/foo" > paths. super_name "/foo" returns an empty string, not "." as I > expected (and even if it returned "." that would be wrong). > > I will send a patch to make super_name work with "/foo" paths after > latest David's FileName patches hit the repo. > > Do I understand it correctly that after 2.1.0 release we are going to > get rid of FileName module (that one that is not in Darcs/Patch)? Are > we going to use System.FilePath or smth else? One thing that hinders darcs adoption of System.FilePath is that we have no assurance of where our homegrown ad-hoc solution matches that of the now de facto solution (System.FilePath) and no one likes the thought of regressions. This of course is a huge hint that someone should sit down with both modules, QuickCheck and HUnit, and then write code that tests the common and corner cases to see where the two modules diverge. If someone did this work, we'd going into a transition with quite a bit more knowldge and also confidence. One thing we have to be cautious about is that we continue to interpret the paths stored in patches the same way we always have. Otherwise, applying patches with a copy of darcs, post refactor, could lead to corrupt repositories. This is a worst case scenario, but I think it's why we're hesitant to jump on the de facto wagon. Jason
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
