Fix test cases #94
Fix test cases #94
Conversation
When there are two conflicting processes, ignore the clocks and always look at the current node topology to resolve the conflict. Fixes the integration test which sometimes did not pass.
This way, the child nodes are not accidentally blacklisted themselves.
README.md
Outdated
| @@ -324,6 +324,17 @@ end | |||
|
|
|||
| MIT | |||
|
|
|||
| ## Testing | |||
|
|
|||
| `mix test` runs a variety of tests, most of the use a cluster of | |||
There was a problem hiding this comment.
There's a minor typo: (...) most of them (...).
|
Any movement on this @bitwalker? |
|
Looks like this conflicts with #85 - is this intended to replace that PR? I can undo the merge and merge this instead if it supercedes it - otherwise please rebase :) |
|
This is indeed intended to replace #85. I see that it has conflicts now, I'll fix those next week, I'm on holiday currently. |
|
Seems most conflicts are related to formatting. |
|
@arjan I did the merge for you, can you verify when you get a chance? I'm pretty sure it was all formatting related, but there were larger changes in one of the test files, and I chose to just favor the changes in this PR rather than try to parse through what all of the changes were, since formatting was mixed in there. Speaking of which, please don't make formatting changes as part of PRs like this, I'm totally fine with those being made in their own PR, or alongside small changes, because they can basically be merged straight away, but it makes large changes harder to review when mixed together, and in cases like this, can result in somewhat painful merges. No worries this time though :) |
|
Thanks so much! |
This has become quite a large commit. I started from #74 to see if I could fix all the test cases. So basically this commit introduces a few things:
mix formatthe codebase (was not really needed, but makes it easier for contributions in the future):swarmapplication inbetween tests, clearing the registry, adjusting the node blacklist, etc.quorum_test.exs; after tracing it down to the usage of the Clock CRDT, I merged Fixed handling of interval tree clocks #85 into this branch, which fixed the remaining tests.SWARM_DEBUG=trueenvironment variable to see all verbose messages (documented in the README).--seed 0argument, to ensure tests can be executed independently of each other.I can imagine this is a big PR to review but I did not touch any logic code except for merging in the #85 branch. (well, except e793c78, but those changes were overwritten by the #85 merge anyway)
Eventually, a lot of (clustering) boilerplate could be lifted out of the test code but I noticed you're already working on that ;-)