When validating a new descriptor against our rend cache failure, we were
added the failure entry to the new cache entry without duplicating. It was
then freed just after the validation ending up in a very bad memory state
that was making tor abort(). To fix this, a dup function has been added and
used just before adding the failure entry.

Fixes #17041

Signed-off-by: David Goulet <dgoulet@ev0ke.net>

Some slower configurations, like OS X i386, need 35 seconds to reliably
bootstrap larger chutney networks. Increase default bootstrap time in
src/test/test-network.h to 35 seconds.

Some platforms have true at different locations, like /usr/bin/true.

No functional changes, but since NoKeepAliveIsolateSOCKSAuth isn't
part of isoflag, it should be checked where all other similar options
are, and bypass the (no-op) masking at the end.

Increase default boostrap time in test-network.sh to 30 seconds,
for larger networks like bridges+ipv6+hs.

This avoids the failure-hiding issues inherent in the retry approach
in #16952.

Yawning Angel authored 9 years ago and Nick Mathewson committed 9 years ago

This controls the circuit dirtyness reset behavior added for Tor
Browser's user experience fix (#15482). Unlike previous iterations
of this patch, the tunable actually works, and is documented.

(These inputs are possible when Shadow starts the world at time_t 0,
and breaks our assumption that Tor didn't exist in the 1970s.)

Fixes regression introduced in 241e6b09. Fixes #16980.

make test-network-all is Makefile target which verifies a series
of test networks generated using test-network.sh and chutney.

It runs IPv6 and mixed version test networks if the prerequisites are
available.

Each test network reports PASS, FAIL, or SKIP.
Closes ticket 16953. Patch by "teor".

Also adds "--hs-multi-client 1" option to TEST_NETWORK_FLAGS.
This resolves #17012.

Larger networks, such as bridges+hs, may fail until #16952 is merged.

Apparently this only happens with clang (or with some particular
clang versions), and only on i386.

Fixes 16970; bug not in any released Tor.

Found by Teor; fix from Yawning.

I think this may fix some bugs with windows exit codes being screwy.

Fixes #15963

Signed-off-by: David Goulet <dgoulet@ev0ke.net>

David Goulet authored 9 years ago and Nick Mathewson committed 9 years ago

In a nutshell, since a circuit can not exit at its entry point, it's very
easy for an attacker to find the hidden service guard if only one EntryNodes
is specified since for that guard, the HS will refuse to build a rendezvous
circuit to it.

For now, the best solution is to stop tor to allow a single EntryNodes for
an hidden service.

Fixes #14917

Signed-off-by: David Goulet <dgoulet@ev0ke.net>

Fixes 16965

(found thanks to teor's chutney haxx)