We can verify a hidden service using the existing chutney.Traffic.TrafficTester as long as the hidden service redirects requests to the TrafficTester's LISTEN_PORT.
I am working on this at the moment, but I still need to verify the code does what I think it should.
Trac: Status: new to assigned Owner: nickm to teor
chutney can now verify hidden services and bridge clients.
But HSs take around 40-45 seconds to bootstrap, even after applying the changes in #13401 (moved).
BCs also take around 35 seconds to bootstrap.
This is compared to basic-min's 10 seconds (an ideal bootstrap time).
I fear this may be a repeat of the timing-debugging-monstrosity that was #13718 (moved).
#13401 (moved) reduces HS bootstrap times to 25 seconds, and contains the corresponding tor and chutney changes.
Unfortunately, bridge client bootstrap times are still around 35 seconds. I think we should be able to adjust these to 15-25 seconds. They can hardly be more complex than hidden services, can they?
Oops, let's do this the other way. Further bootstrap time improvements to #14034 (moved), the rest is ready for review in 0.2.6, branches in #13401 (moved).
Trac: Status: assigned to needs_review Milestone: Tor: 0.2.7.x-final to Tor: 0.2.6.x-final
This means that tor's src/test/test-network.sh will bootstrap reliably with --flavour bridges+ipv6+hs with the default delay of 25 seconds.
This will verify data transmission from:
every client to an arbitrary exit,
every bridge client to an arbitrary exit, and
an arbitrary client or bridge client to every hidden service
Which paves the way for chutney-based performance testing in #14175 (moved).