In preparation for the kvdb->migration code, this commit updates
`insertChannel` to return the ID of the newly inserted channel along
with the IDs of the nodes that the channel links to.
Refactor channelIDToBytes to return a slice instead of an 8 byte array
so that we dont need to use `[:]` everywhere.
Also make sure we are using this helper everywhere.
Create an abstract hashAccumulator interface for the Channel graph
bootstrapper so that we can later introduce a deterministic accumulator
to be used during testing.
Here we start using the newly added message_flags and channel_flags
columns of the channel_policies table. The test added previoulsy to
demonstrate the bug is now updated to show that the bug has been fixed.
We need to explicitly store the entire bitfield types since we may have
channel_updates with bitfields containing bits we just dont need or
understand but we still need to store the entire bitfield so that the
reconstructed announcement remains valid.
This commit only adds the new columns but does not use them yet. NOTE:
this is ok since the migration adding this schema is not available in
the production build yet.
This commit adds a new graph store unit test that demonstrates that
there is currently a bug in the SQL version of the graph store where
unknown message or channel flags in a channel_update are not persisted
correctly. This will be fixed in an upcoming commit.
This commit demonstrates that currently the Addresses field of a
LightningNode is not populated correctly if the node has no addresses if
the SQL graph store is being used. This will be fixed in an upcoming
commit.
If a method returns an error, we should assume all other parameters to
be nil unless the documentation explicitly says otherwise. So here, we
fix a log line where a dereference is made to an object that will be nil
due to an error being returned.
The comment of FetchChannelEdgesByID says that if the ErrZombieEdge
error is returned, then the ChannelEdgeInfo return parameter will also
still be populated and returned (ie, wont be nil). This commit updates
the SQLStore implementation of FetchChannelEdgesByID to do this. This is
needed to prevent nil dereference panics at any call-sites that depend
on the method working as it describes.
On startup, we currently blindly overwrite our node announcement info
for our source node. This includes overwriting the existing LastUpdate
timestamp. This can cause an issue with our new SQL backends in the case
that the new timestamp is not greater than the previously written
timestamp. So here, we first fetch the source node and make sure to use
a timestamp that is greater than the previosly set one.
Replace all usages of the "github.com/go-errors/errors" and
"github.com/pkg/errors" packages with the standard lib's "errors"
package. This ensures that error wrapping and `errors.Is` checks will
work as expected.
In this commit, a new `test_native_sql` build flag is defined. If this
build flag is used along with the `--use-native-sql` config option, then
the SQLStore implementation of the graphdb.V1Store will be initialised.
This is then used to run our itest suite against the new SQLStore graph
implementation.
NOTE that this only works for new nodes currently - the migration from
kv-to-sql is yet to be implemeneted.
And delete the old MakeTestGraph function.
Here we need to update some of our unit tests to ensure that any updates
to existing nodes or channel policies have newer timestamps. This is so
that we don't violate our SQL DB constraints that prevent updates to a
node or channel policy record if the new update is not newer (has a
newer LastUpdate time) than the currently presisted record).
