[MERGE][1.6][#254797] Fix NotImplementedError when probing for smart protocol via HTTP.

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Andrew Bennetts wrote:
> John Arbash Meinel wrote:
> [...]

...

>> The test also seems a bit indirect. I realize it probably would fail to
>> BzrDir.open('smart+http://') but it doesn't really feel like you are testing
>> that Medium's are implementing the appropriate interface.
> 
> Yes, it would be good to have per-medium tests.  That's on a todo list of mine
> somewhere...
> 
>> Anyway, this seems like a minimal fix for 1.6, and there isn't anything I
>> would strictly block on. And it mostly comes back to wanting
>> "smart_medium_implementations/" testing.
> 
> Right, exactly :)
> 
> It would be good to find time to work on those tests, although I think the
> medium layer is pretty settled now (despite this bug), so it's hard to feel like
> it's worth doing before other bugs and features.  If I do find myself doing
> non-trivial work in it I will try to make the time to do implementations tests
> first, though.
> 

Well, this is at least the 2nd (3rd?) non-trivial implementation mismatch. We
had "_server_is_at_leastXXX" and now "_read_line" (are there others?). I'd
have to say after 2 fatal regressions, it is time to do something real about it.

Personally, I would rather see it before the effort testing, even though one
is "covering over possible bugs" and the other is "implementing new features".

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFIpDwMJdeBCYSNAAMRAih/AKDDkt/l7SpSUFkaLbAWtUD5CRgQMgCeKHVN
FeHp+dpYFDnB0xzbOhBzJO4=
=X7hj
-----END PGP SIGNATURE-----