just skip any unparseable rbr events

[bbeaudreault]

@alainjobart not sure if this is what you had in mind. I kept the panic for the NYI types, since I figured that may have other consequences.

[alainjobart]

That works, thanks!

[alainjobart]

Ah you also have to fix go/mysqlconn/replication_test.go for this to work, see travis.

[alainjobart]

I'll merge when travis passes. :)

[bbeaudreault]

@alainjobart can you kick the test off again? i think the failure looks like our usual unrelated problem?

[michael-berlin]

@alainjobart can you kick the test off again? i think the failure looks like our usual unrelated problem?

Out of curiosity: When you log in with your GitHub account on Travis CI, can you then restart the test yourself?

[bbeaudreault]

@michael-berlin I'd love to be able to, but unfortunately I can't. I've looked for it a few times and just looked again. I am logged in with my GH account. I think possibly only committers can.

[michael-berlin]

I've added you as collaborator to our project. Please accept the invite in your email.

Maybe after that you're able to retrigger tests yourself.

The other thing to keep in mind: Some tests are known to be flaky e.g. the "local_example" when it cannot download all artifacts from Maven Central. When these fail, I ignore it and merge it anyway. We can still fix forward if only that test was broken. If you see a flaky test a lot of times, feel free to file an issue. We should fix that to get the flaky rate down.

[michael-berlin]

Forgot to mention: As a collaborator, you're able to push branches to our repo, but not to the master or release branches. That means merges into master still have to be done by the team. Write access to the repo is probably not relevant for you since you're working in your own fork.

[bbeaudreault]

Thanks @michael-berlin. I appreciate the effort, but sadly that didn't seem to work. A quick googling shows that only those with write access see the refresh button that kicks off a build. I can probably make due with pushing an empty commit or amending and force pushing as a way to kick off new builds for now.

[michael-berlin]

Yeah, that's too bad.

I agree: Force pushing or not worrying about known flaky tests are good workarounds.

[bbeaudreault]

@michael-berlin FYI not sure what changed (maybe it was cached when we were testing this before), but I can now restart individual jobs in a build or entire builds, or cancel a running build. So either you did something else, or collaborator role eventually makes this functionality available. Thanks again!