close/reopen

A bunch of tests timed out after 72 hours (so they claimed). Not sure what was going on with that.

Something is off.

The skip for azure did not work
Neither did the skip for shippable (which, it turns out, will be turned off in May 2021)
It also seems the github actions CI ran.

Neither did the skip for shippable (which, it turns out, will be turned off in May 2021)

Oh fun, more CI migration. Guess we're going to Drone for ARM 64 then. Let's not do it before it's necessary, the landscape may have changed again by then.

The skip for azure did not work

This is expected, see the docs in the PR.

Neither did the skip for shippable

There was no skip for shippable, so this is also as expected.

It also seems the github actions CI ran.

This I did not expect, will look into it.

Oh fun, more CI migration

I did see a jFrog (new owner of Shippable) app for Github actions in the marketplace. jFrog seems to supply an interface to AWS, Azure, or Google, but I'm not clear on how it should be used.

@rgommers did you figure out why github actions ran?

@rgommers did you figure out why github actions ran?

I didn't, no time to look into it.

We removed Shippable from the CI testing and now rely on TravisCI for arm64 testing.

+1 Thanks Ralf.

(I haven't merged because I'm not sure what caused the CirceCI checks to fail.)

Looks like the CircleCI debian image needs to be updated? Not sure why this is only showing up here, though.

Probably some ancient issue, I just rebased on main, lets see.

I guess we still have the question of why the github actions tests and the Travis CI tests ran for this PR even though a commit includes [skip github] and [skip travis].

I tried adding

[skip github]
[skip travis]

to the second commit as well at least, lets see if it does something (the docs say azure won't work anyway).

I have to admit, it would be a lot easier if we could have something like [ci skip docs], [ci docs only] and maybe [ci python only] (the latter to skip travis jobs for weird architectures and maybe debug builds when the change is a very simple python change).

But, I suppose that is just dreaming :).

Thanks @rgommers. This documents how we hope it works, we can iterate in the future to make it more accurate if something is off.