Solution for baz2bzr revision ids on windows

[Martin Pool]

On 11 Jan 2006, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
> John Arbash Meinel wrote:
> > I would have preferred to have bzr handle some of this internally. But
> > depending on how we do it, bzr might end up doing the same thing, so we
> > can just stop translating in baz2bzr if we know that bzr is going to
> > translate itself.
> 
> Because these ids will be stored in revision and inventory files
> directly, I believe that would require another namespace upgrade.

Aaron/Robert's point that letting clients not worry about what
characters are safe in ids is persuasive, so I think in the future the
storage should escape them.  However, that probably requires a format
change.  

(I suppose we could abuse the current format and say files can be named
with either the escaped or unescaped forms.  It's a bit ugly but it
could do.)

> > So if I'm understanding you correctly, you are willing to create
> > 'Arch-2' which is a namespace which passes
> > urllib.quote(x.encode('utf-8')) as the revision-id.
> 
> Right.

Perhaps this is the best short-term option for you?  On the other hand
it will make it incompatible with other imports.

So I think we need a format bump that does this escaping in the stores.
There is some risk in adding a new format without proper testing of
conversion between formats, and that operations work on all supported
formats.  But I suppose you can always do such a format on a separate
branch and us it yourself.  The actual changes should be
straightforward; it's just the test infrastructure that will take a bit
of work.  But Robert and I are looking at that now.

-- 
Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060112/e9794f12/attachment.pgp