On Tue, Feb 17, 2004 at 04:06:28PM -0500, Christopher Faylor wrote: > On Tue, Feb 17, 2004 at 03:56:52PM -0500, Pierre A. Humblet wrote:
> >Good plan. After looking into it, I now think we could get rid of the > >normalized path storage entirely. It's mostly used in get_name(), which is > >90% for printf(). A large part of the remaining 10% could use > >get_win32_name instead (and be faster). A few oddball cases need to be > >looked into. > > Using the win32_name causes strange side effects. A win32 name is not > always equivalent to a unix name. You lose state. I just fixed a > problem with this in stat recently. I initially fixed it by using the > unix name but then correctly fixed it by passing the fhandler pointer > around since that ended up doing many operations only once. > > Eventually, moving everything to use fhandler_* could get rid of a > requirement to store names at all, although we probably still want to > keep the win32 name. Obviously you have thought about it more than I have! I had in mind the case of the fhandler_*. There are only two other cases where PC_POSIX is used in path_conv anyway, so they can be taken care of individually. I agree that win32 names will remain needed. When you have a minute or two, could you think of an example where the win32 name would have problems replacing the unix name in fhandler_*? Pierre -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
