On Wed, Oct 26, 2005 at 23:56:51 -0700, Matt Mackall wrote:
> On Thu, Oct 27, 2005 at 08:48:12AM +0200, Jan Hudec wrote:
> > On Wed, Oct 26, 2005 at 16:41:18 -0700, Matt Mackall wrote:
> > > On Thu, Oct 27, 2005 at 12:17:16AM +0200, Jan Hudec wrote:
> > > > Hello All,
> > > >
> > > > I found some more points that confuse me. Specifically, what are the
> > > > rules
> > > > for paths within dirstate.walkhelper? Because on one hand the files are
> > > > opened using that paths, which means they must be either absolute or
> > > > relative
> > > > to current directory. But on the other hand they are looked up the
> > > > dirstate
> > > > map, where they can hardly be relative to current working directory,
> > > > which is
> > > > completely arbitrary. The code where I want to use it always passes full
> > > > paths and changing working directory is STRONGLY undesirable (I know
> > > > how to
> > > > make changing directory work, but it's hacky).
> > >
> > > Paths handled anywhere beneath the command layer should be full paths.
> > > Filters, on the other hand, may be relative.
> >
> > A bit more specifically, if I have a repo in /path/to/repo and a file
> > /path/to/repo/subdir/file (and working directory is /somewhere/in/hell),
> > what is in the dirstate.map?
> > - subdir/file
>
> This one.
Thanks. So it's repository relative. I thought that, but I wanted to be
sure.
> > - ./subdir/file
> > - /path/to/repo/subdir/file
> >
> > As I've seen the paths in dirstate.map are always '/'-separated, even if
> > they are '\'-separated natively, but what will I get in the statmatch
> > callback in dirstate.walkhelper for the above case on Windows?
> > - subdir/file
>
> And should be this one.
Well, I ask because the statmatch defined in dirstate.changes calls
util.pconvert on it's argument first.
--
Jan 'Bulb' Hudec <bulb@xxxxxx>
signature.asc
Description: Digital signature
|
