Microdescriptor search in #33428 is almost finished, so change the
comment to refer to onion service descriptor search in #33609.

If there is a time calculation bug, or a large clock skew, fail
bootstrap.

Diagnostics for macOS Travis hangs.

chutney was calculating some elapsed times incorrectly, leading to longer
wait times for network bootstrap.

Bugfix on 33615.

chutney was displaying any node set without directory authorities as
"caches and clients".

Now, chutney only displays "caches and clients" if the node set contains
all caches and clients, and no dir auths.

Bugfix on 33615.

Bugfix on 33615.

Some legacy Tor versions (0.3.5 and earlier) have microdescriptor
download bugs. Chutney triggers those bugs by verfying early, and
rapidly reconnecting to the SOCKSPort (#33598).

So chutney waits for a minimum amount of time after launching the
network (65 seconds on 0.3.5 and earlier), regardless of bootstrap
status or directory info distribution.

We also change the unchecked descriptor wait time to:
  * 10 seconds for microdesc downloads, and
  * 30 seconds for onion services to publish their descriptors.

Fixes bug 33615.

Preparation for 33615.

Documentation for 33379.

Cleanup after 33361.

Previously, the script reset the number of failures to 1.

Fix on 33583.

Set CHUTNEY_ALLOW_FAILURES to 2, so that CI tolerates a single failure
during each test-network.sh run.

Part of 33583.

We want to make sure tor has fully exited, and the OS has reclaimed
its resources.

The pause helps if the failure happened due to system load.

Part of 33583.

The current script output makes it easy to confuse bootstrap attempts
and verify rounds.

Change the output to make it clearer.

Part of 33583.

Relays and bridges will now perform reachability self-checks, before
publishing their descriptors.

Continue to set "AssumeReachable 1" on authorities, to stop them doing
authority to relay/bridge reachability checks. (These checks happen too
slowly for chutney's test networks.)

Part of 33583.

Wait for a consistent amount of time between verify attempts,
regardless of the time elapsed during the attempt.

Try to make at least 2 verify attempts per consensus interval,
but also try to limit the number of attempts.

Also increase the default timeout: sometimes 3 seconds is too
short on a heavily-loaded machine.

These changes should make verification more reliable.

Part of 33379.

Tor 0.3.5 is unstable with exit relays in mixed+hs-v2, so we use
non-exit relays.

(Recent tor versions use mixed+hs-v23, and they support exit relays.)

Part of 33379.

Authorities and relays need 3 *other* authorities/relays to bootstrap.
And Tor 0.3.5 needs 5 authorities/relays in the consensus.

We add authorities, because some networks only have one relay or exit,
to make sure all paths go through that relay/exit.

Part of 33232.

Avoids a race condition in microdescriptor downloads in 0.3.5.

Once we check for microdescriptors in 33428, we should be able to reduce
this timeout.

Part of 33379.

Bridges with AssumeReachable fail to publish their descriptors, because
they try very early, when they haven't bootstrapped.

Turning off AssumeReachable in 33232 should fix this issue.

This changes makes these chutney networks succeed:
* bridges-min
* bridges+ipv6-min

Part of 33379.

Stop expecting bridges to be published in the consensus.

Also, stop expecting bridge clients to:
* have bridge microdescriptors, and
* have consensus relay full descriptors.

Partially fixes these chutney networks:
* bridges-min
* bridges+ipv6-min

Part of 33379.

Check that every relay has published its directory info to the
authorities, it is in the consensus, and every node has downloaded it.

Check that every bridge has published its directory info to the
bridge authority, and it is in the bridge networkstatus.

Check that every bridge client has downloaded bridge descriptors.

This code currently does not work for these chutney networks:
* bridges-min
* bridges+ipv6-min

All other networks pass the new tests.

Part of 33379.

Part of 33379.

And update some old code to use them.

Part of 33379.

Part of 33379.

When $CHUTNEY_WARNINGS_SKIP is true, skip all warnings.

Part of 33379.

Before launching "chutney verify", test-network checks that all the
nodes have bootstrapped.

Part of 33232.

wait_for_bootstrap now prints bootstrap progress every 20 seconds.

Part of 33232.

Split print_bootstrap_status() out into its own function.

Part of 33232.

The previous code counted the number of sleeps.

Using wallclock time is more accurate, and the code is simpler.

Part of 33232.

Part of 33232.

These warnings are caused by known bugs in tor.
But they confuse some chutney users.

Related to tor ticket 33354, and similar tickets.

Chutney creates a lot of late votes, because its consensus intervals
are very short. So we want to ignore those late vote warnings by
default.

We will still see any consensus failure warnings. And the full logs
will still be available for diagnostics.

Related to tor's ticket 4631.

Unfortunately, we use the UK English spelling for this variable.

Obviously correct changes to already-reviewed code.

This network was running chutney's default network flavour before,
changing to basic-min shouldn't make any difference.

Obviously correct changes to already-reviewed code.