Re: [Monotone-devel] [PATCH] fix use of hostname from parsed uri

[Roland McGrath]

> Anyway, I'll find time this weekend to merge your patch and some tests for it.

Thanks very much.

> (Did it seriously take all day to figure out the test suite?  I know
> it's not well documented but I didn't think it could be *that*
> confusing.)

I didn't spend all day!  I haven't figured it out!

> > I can't figure out why tests/netsync_mtn_uri_scheme doesn't get run
> > already, or if it does, why it doesn't already fail the same way that I see
> > when I try any simple manual use of a mtn:// uri.
> 
> Hmm.  Did you try urls of the form mtn://host?branch instead of
> mtn://host/branch ?  That's the only thing I can think of.

I tried by hand what appear to be the cases that test does and they don't
work either.  In no example anywhere (mine or anyone's) is mtn://host/branch
a proper syntax.  I mentioned mtn://host/name, where "name" is entirely
something to be considered by usher.

> You might find the debian/monotone-server.* scripts helpful, although
> I have no idea how much of that is applicable to Redhat-style systems
> or to usher setups.  It would be great to make that stuff less
> distro-specific.

The Fedora monotone-server package already does the equivalent to that.
(I'm not sure which of us did it first, in fact.)  That is a setup for
running "mtn serve" from one db.  What I'm talking about, as a sensible
hosting setup that won't be laughed out of town by anyone running a hosting
service for SCMs, requires usher (i.e., many separate db's).


Thanks,
Roland