We had a couple of instances where a plugin would be killed by `lightningd`
because we were returning a result of an exception twice, and it was hard to
trace down the logic error in the user plugin that caused that. This patch
adds a traceback the first time we return a result/exception, and raise an
exception with a stacktrace of the first termination when a second one comes
in.
This can still terminate the plugin, but the programmer gets a clear
indication where the result was set, and can potentially even recover from it.
Changelog-Added: pyln: Plugin method and hook requests prevent the plugin developer from accidentally setting the result multiple times, and will raise an exception detailing where the result was first set.
It is often pretty usefuk to use the builtin logging module to debug things,
including libraries that a plugin may use. This adds a simple
`PluginLogHandler` that maps the python logging levels to the `lightningd`
logging levels, and formats the record in a way that it doesn't clutter up the
`lightningd` logs (no duplicate timestamps and levels).
This allow us to tweak the log level that is reported to `lightningd` simply
using the following
```python3
import logging
logging.basicConfig(level=logging.DEBUG)
```
Notice that in order for the logs to be displayed on the terminal or the
logfile, both the logging level in the plugin _and_ the `--log-level`
`lightningd` is running need to be adjusted (the python logging level only
controls which messages get forwarded to `lightningd`, it does not have the
power to overrule `lightningd` about what to actually display).
I chose `logging.INFO` as the default, since libraries have a tendency to spew
out everything in `logging.DEBUG` mode
Changelog-Added: pyln: Plugins have been integrated with the `logging` module for easier debugging and error reporting.
Since we start a new instance of postgres for each test we may end up swamped
and the startup can take a bit longer. So let's loop until we get a success.
It was really flaky, especially under `test_mpp_interference_2`, most likely
due to multiple calls to `fund_channel`. This commit looks for the specific
txids in the `listfunds` output and the `getrawmempool` output, avoiding
strange artifacts from multiple calls.
For performance reasons we were starting one for each session, which caused
the same postgres DB to be re-used for multiple tests (all test run in the
same worker process), but this could lead to interactions if there is a
timeout or a test happens to touch the `db_provider`. It turns out that we
were only saving about 15 seconds on a 1250 second run anyway, which is a
small cost for increased test isolation.
We were not removing the base test directory if we had other files in there,
which was the case for postgres runs. This now explicitly check for `test_*`
directories which are an indicator of a failed test.
We had a couple of issues with workers dying and attempting to re-initialize
the database while it was already initialized. This will look for a free
directory and just start the DB in there, allowing workers to be better
isolated.
Several times we had issues with plugins not being able to re-encode an RPC
result because they forgot to use the custom encoder class. This allows us to
patch the JSONEncoder when we start the RPC or the plugin and automagically
support classes that provide a `to_json` method.
The next patch perturbed things enough that we suddenly started
getting (with --track-origins=yes):
Valgrind error file: valgrind-errors.120470
==120470== Use of uninitialised value of size 8
==120470== at 0x14EBD5: htable_val (htable.c:150)
==120470== by 0x14EC3C: htable_firstval_ (htable.c:165)
==120470== by 0x14F583: htable_del_ (htable.c:349)
==120470== by 0x11825D: pointer_referenced (memleak.c:65)
==120470== by 0x118485: scan_for_pointers (memleak.c:121)
==120470== by 0x118500: memleak_remove_region (memleak.c:130)
==120470== by 0x118A30: call_memleak_helpers (memleak.c:257)
==120470== by 0x118A8B: call_memleak_helpers (memleak.c:262)
==120470== by 0x118A8B: call_memleak_helpers (memleak.c:262)
==120470== by 0x118B25: memleak_find_allocations (memleak.c:278)
==120470== by 0x10EB12: closing_dev_memleak (closingd.c:584)
==120470== by 0x10F3E2: main (closingd.c:783)
==120470== Uninitialised value was created by a heap allocation
==120470== at 0x483B7F3: malloc (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
==120470== by 0x1604E8: allocate (tal.c:250)
==120470== by 0x160AA9: tal_alloc_ (tal.c:428)
==120470== by 0x119BE0: new_per_peer_state (per_peer_state.c:24)
==120470== by 0x11A101: fromwire_per_peer_state (per_peer_state.c:95)
==120470== by 0x10FB7C: fromwire_closingd_init (closingd_wiregen.c:103)
==120470== by 0x10ED15: main (closingd.c:626)
==120470==
This is because there is uninitialized padding at the end of struct
peer_state.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Currently ``LightningConnection.remote_pubkey`` is set to ``None`` if the node is not the handshake initiator. This sets it to ``rs`` in act three from the receiver side
Since pyln-bolt* specify the 0.8.4 version which we didn't upload, and the
requirements.txt specify ==0.8.4, we need to backfill that version, even if we
could just bump it directly to 0.9.1.
Hooks do not tolerate failures at all. If we return a JSON-RPC error to a hook
call the only thing the main daemon can really do is to crash. This commit
adds a mapping of error to a safe fallback result, including a warning to the
node operator that this should be addressed in the plugin. The warning is
reported as a `**BROKEN**` message, and should therefore fail any testing done
on the plugin.
Changelog-Fixed: pyln: Fixed HTLCs hanging indefinitely if the hook function raises an exception. A safe fallback result is now returned instead.
Fixes: #2679
Changelog-Added: JSON-RPC: New `multiwithdraw` command to batch multiple onchain sends in a single transaction. Note it shuffles inputs and outputs, does not use BIP69.
Re-write start_ln such that we can create up to 10 nodes locally for
testing. Useful for scenarios where more than two nodes are needed
Changelog-Changed: contrib: startup_regtest.sh `startup_ln` now takes a number of nodes to create as a parameter
This is an extra safety check for dual funding, where we only want to sign
the inputs we provided!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Added: JSON-RPC: `signpsbt` takes an optional `signonly` array to limit what inputs to sign.
Required for dual funding where the opener sets it.
Changelog-Added: JSON-RPC: `fundpsbt` takes a new `locktime` parameter
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This allows plugins to choose how to present things in getmanifest.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Added: plugins: `getmanifest` may now include "allow-deprecated-apis" boolean flag.
Changelog-Deprecated: plugins: `getmanifest` without any parameters; plugins should accept any parameters for future use.
And when it's set, and we're SLOW_MACHINE, simply disable valgrind.
Since Travis (SLOW_MACHINE=1) only does VALGRIND=1 DEVELOPER=1 tests,
and VALGRIND=0 DEVELOPER=0 tests, it was missing tests which needed
DEVELOPER and !VALGRIND.
Instead, this demotes them to non-valgrind tests for SLOW_MACHINEs.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We can query all the txids at once, rather than one at a time.
Doesn't make any measurable difference to full runtime testing here
though.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
We might have funds prior to calling join_nodes(), so testing that
we've all seen the block is better.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This is what fund_channel() does, which is more thorough than what
we were doing. But since the order of the logs is undefined, we need
to be a little careful.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Reduces VALGRIND=1 node_factory.line_graph(5) time on my laptop from 42s to 36s.
This is simply because forking all the subdaemons just to check the
version is very expensive under valgrind.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
fundpsbt forces the caller to manually add their weight * feerate
to the satoshis they ask for. That means no named feerates.
Instead, create a startweight parameter and do the calc for them
internally, and return the feerate we used (and, while we're at it,
the estimated final weight).
This API change is best done now, as it would otherwise have to
be appended as a parameter.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Added: JSON-RPC: new call `signpsbt` which will add the wallet's signatures to a provided psbt
Changelog-Added: JSON-RPC: new call `sendpsbt` which will finalize and send a signed PSBT
Reserve and unreserve wallet UTXOs using a PSBT which includes those
inputs.
Note that currently we unreserve inputs everytime the node restarts.
This will be addressed in a future commit.
Changelog-Added: JSON-RPC: Adds two new rpc methods, `reserveinputs` and `unreserveinputs`, which allow for reserving or unreserving wallet UTXOs
Now they look like 1.0.1.137, so you can explicitly depend on a csv change
(without caring about a textual change).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
1. version was 0.0.2 in setup.py, which means we didn't get the dist/ files we expected.
2. We need 'bdist_wheel' to make the .whl file.
3. --no-site-packaged was apparently removed in 0.20.0, and was default long before that.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This includes some real bugfixes, since it noticed some places we were
being loose with different types!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
They're almost entirely autogenerated, and we use symlinks into the
top directory to reduce replication.
They can't be under pyln.spec.message, because a package can't also
be a namespace.
We also add fulltext and desc fields, and exclude our "gen" files from
flake8, since the spec quotes contain weird whitespace.
Changelog-Added: Python: pyln.spec.bolt{1,2,4,7} packages.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Remove non-existant pyln.proto.bolts. bolts will have separate setup.py, so we
can rev the versions individually.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This delivers the message contents in a much friendlier form for
manipulation: in particular, it makes it easy to compare two
messages without having to know all the message type internals.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
If they don't exist, that's OK. These will eventually be going away
from the spec, but there are still some in gossip messages for now.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Exposing the array types is required for our dummyrunner in the lnprototest suite, since
it wants to be able to generate fake fields.
The set_field is similarly useful.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Other changes along the way:
1. In a couple of places we passed None as a dummy for for
`otherfields` where {} is just as good.
2. Turned bytes into hex for errors.
3. Remove nonsensical (unused) get_tlv_by_number() function from MessageNamespace
4. Renamed unrelated-but-overlapping `field_from_csv` and
`type_from_csv` static methods, since mypy thought they should have
the same type.
5. Unknown tlv fields are placed in dict as strings, not ints, for
type simplicity.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
These are autogenerated, but now they export their own
MessageNamespace, as well as the raw csv.
They also expose their SubtypeTypes, MessageTypes and TlvStreamTypes,
though in theory these could clash (they don't for now, and it'd be
kinda awkward if they did).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Christian Decker
Rusty Russell
fiatjaf
Antoine Poinsot
niftynei
Sergi Delgado Segura
Michael Schmoock
ZmnSCPxj jxPCSnmZ
nicolas.dorier