Re: [Gnu-arch-users] Re: [MERGE REQUEST] unify various commands around arch_fqvsn_from_tree_and_input

[Robert Collins]

On Thu, 2004-05-20 at 12:50, Miles Bader wrote:
> Aaron Bentley <address@hidden> writes:
> > Yeah, it's ugly.  Valid revision names are always considered to be 
> > revisions, but I believe it's possible to express any file path so that 
> > it's not a valid revision.  For example, a directory named 
> > "address@hidden/cat--branch--1.0--patch-3" could be referred to as 
> > "./address@hidden/cat--branch--1.0--patch-3", which isn't valid.
> 
> Yeah; I would be perfectly happy if using directory names required a
> magic syntax.  E.g., for doing a delta against the directory FOO,
> something really magic like `dir:FOO' (which has the advantage that it's
> clearly not a normal filename), or `./FOO' (which is somewhat compatible
> -- we'd probably want to accept just `.' as a special case for the
> latter too).
> 
> Of course there's always that annoying backward compatibility thing...

I don't see the issue: the grammar is not ambiguous: first it tries
archive paths, then file paths. If you have a file path which is also a
archive path, just prepend ./ to it, which is not valid for any
component of an archive path.

Rob
-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.

signature.asc
Description: This is a digitally signed message part