Since we seem to have some isolation concerns when re-generating the
same HSM secret and re-parsing the blockchain some blocks in the past.
This also alleviates the problem of printing to a logging stream that
has been closed. Previously bitcoind would keep running despite a test
had failed and continue logging to the, now closed, StringIO that
py.test uses when capturing stdout.
The performance impact seems to be 1-3 second per test, not too bad
IMHO for increased test isolation and cleaner logs:
|--------------------+---------------+----------|
| | No_valgrind | Valgrind |
|--------------------+---------------+----------|
| bitcoind per suite | 10 min 24 sec | 46:15.31 |
| bitcoind per test | 11 min 38 sec | 49:21.64 |
|--------------------+---------------+----------|
Signed-off-by: Christian Decker <decker.christian@gmail.com>
I was examining a test_onchain_timeout failure, and realized that we
were forgetting a peer even though we'd just spent the HTLC_TIMEOUT_TX!
This reveals that we weren't resolving an output when we stole the preimage
from it, like we should.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Since we panic when we see our root reorg out, even if we're not doing
anything yet, restoring the 100 block margin is the simplest fix.
Unfortunately this means adding a 100-block spacer in the tests, so things
don't get confused.
Fixes: #511
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We stopped too early; we should continue and make sure it all goes well.
This means we have to fix them to be deterministic: by generating 2
blocks at once in test_htlc_in_timeout, we raced between fulfill and
timeout on the HTLC. Now it's always fulfilled.
Also, fixed confusing comments: l1 doesn't drop to chain.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
OUR_HTLC_TIMEOUT_TO_US = normal tx, used to timeout htlc in their commit tx.
OUR_HTLC_TIMEOUT_TX = dual-sig tx with delay, used to timeout htlc in our commit tx.
Only one test looks at that string, so fix that too.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
The test_reconnect_normal test is failing rather consistently on 32bit
architectures, disabling to reduce noise. Issue #468 tracks progress.
Signed-off-by: Christian Decker <decker.christian@gmail.com>
Relatively simple: until we reach funding-depth the channels should be
known locally, so we can already route through them, but they should
not be announced to peers to which the connection is non-local.
Signed-off-by: Christian Decker <decker.christian@gmail.com>
If send_htlc_out() fails, it doesn't initialize pc->out; that can
make us think it's still in progress.
Reported-by: Jonas Nick
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
All peers come from gossipd, and maintain an fd to talk to it. Sometimes
we hand the peer back, but to avoid a race, we always recreated it.
The race was that a daemon closed the gossip_fd, which made gossipd
forget the peer, then master handed the peer back to gossipd. We stop
the race by never closing the gossipfd, but hand it back to gossipd
for closing.
Now gossipd has to accept two fds, but the handling of peers is far
clearer.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We should not disconnect from a peer just because it fails opening; we
should return it to gossipd, and give a meaningful error.
Closes: #401
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We don't use it yet, but now we'll decode correctly.
See: https://github.com/lightningnetwork/lightning-rfc/pull/317
lightning-rfc commit: ef053c09431442697ab46e83f9d3f86e3510a18e
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
If you run locally, it fails occasionally; presumably because it
sees previous funds. Use a random HSM key for that teste.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I noticed some breakage with git master:
1. getinfo no longer supported (for us, use getblockchaininfo)
2. generate no longer supported (use generatetoaddress)
Both these options are supported at least in 0.15, too.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This seems to happen when we manage to check between the
channel_announcement and the channel_update being processed.
Signed-off-by: Christian Decker <decker.christian@gmail.com>
Add two simple tests: one for a single direct payment and one with
hundreds of parallel payments, reusing the same route.
Signed-off-by: Christian Decker <decker.christian@gmail.com>
