add regression test for presignature timeout#500
Conversation
Adds a test that would fail without the fix in sig-net#485
|
if things in CI work out the same as the do locally, tests should pass on this PR and fail over at #501 |
jakmeier
force-pushed
the
regression-485
branch
from
12e6d68 to
6a073ff
Compare
| .build() | ||
| .await; | ||
|
|
||
| tokio::time::timeout(Duration::from_secs(10), network.wait_for_presignatures(1)) |
There was a problem hiding this comment.
To be sure I understand this test correctly.
- We start 3 nodes
- They try to generate 5 per-signatures
- Some or all (not clear) of this protocols fail, since we drop 20 per-signature messages on each node
- Nodes start new protocols, at least one of them succeeds.
It means that old protocols are not waiting for new messages forever and the counter of ongoing generators represents protocols that are actually alive.
The requirement for such test is: The number of failed protocols must be >= number of concurrent pre-signature generations. This is not guaranteed by the test itself, but it is very nice to have anyways.
There was a problem hiding this comment.
yeah this is very specific to the bug we had the other week
The number of failed protocols must be >= number of concurrent pre-signature generations.
yes
This is not guaranteed by the test itself, but it is very nice to have anyways.
True, the test relies on the default setup using a small enough max concurrent generators / max introduced setting. Would be nicer to fix it in the test... I might get back to it later if I want to clean things up a bit more but will merge it for now as-is.
2 participants
Adds a test that would fail without the fix in #485