[MERGE][bug #173010] initial branch over http: can show no ui activity for a very long time

Andrew Bennetts andrew at canonical.com
Thu Dec 6 12:46:31 GMT 2007


Andrew Bennetts wrote:
> Vincent Ladeuil wrote:
> > "Gimme my blinkenlights back !" said the user...
> > 
> > Here you are.
> > 
> > The attached fix is a complete rewrite of the http response
> > handling.
> [...]
> 
> This is pretty cool stuff.  I'm going to be a wuss for the moment and make this
> vote:
> 
> bb:comment
> 
> But I intend to change that to something more helpful shortly :)  This is just

Thinking about it, this is pretty good code and I'm definitely comfortable with
this landing in bzr.dev once my comments (and John's) are taken into account,
so I'm updating my vote to:

bb:tweak

I'm still on the fence about if we should risk this for 1.0.  Luckily for me,
Martin is the RM and gets to make the tough decision here rather than me ;)

One further comment: playing with my --coverage hack for selftest, I see there's
a lot of unexercised error handling (i.e. except blocks) in
bzrlib/transport/http/response.py.  I'd really like to see those cases exercised
by the test suite, so that we don't risk giving a user an embarrassing traceback
rather than a friendly error message.  I did a quick comparision versus the
coverage report without your change, and it looks like with your change there
are more unexercised blocks than before, almost all error handling (and also the
whence=1 case I mentioned on IRC and you've already said you're adding a test
for).  It would be good to test that all those code paths are working (or at
least not trivially broken).

-Andrew.




More information about the bazaar mailing list