+1 on retiring WSL1 runners.   Do you anticipate _any_ unique behavior in
WSL2 that we should take into account?

Cheers,

Sander

On Fri, Aug 21, 2020, 21:20 Benjamin Schubert <[email protected]>
wrote:

> Hey everyone,
>
> TLDR: I think we should stop testing on WSL.
>
> History of WSL and our testing on it:
>
> Around a year ago, we introduced WSL runners to test BuildStream, since
> WSL1 had many weird quirks compared to running on a real Linux environment.
> This allowed us to catch quite a few bugs on this platform.
>
> However, the platform has evolved a lot. We now have WSL2, which is a
> Linux VM, running on HyperV, with a real kernel and all we need. I haven't
> found a single bug on WSL2 that I would not have seen on Ubuntu 20.04
> (which is the version my WSL system use). Microsoft recently announced that
> they would release WSL2 to Windows 1903 and 1909 too.
>
> Given this, I don't think there is much of an incentive for users to stay
> on WSL1, and would thus expect most people to move to WSL2.
>
> With the work to move to GitHub and GitHub Actions, it was seen that the
> WSL runners there were quite flaky and thus hard to trust the tests running
> there.
>
> What we would loose:
>
> WSL1 is currently the only test that tests BuildStream on a non-linux
> kernel (being an emulated kernel with not everything present, thus behaving
> more like a generic unix system). So we would lose our current tests around
> unix-ish systems.
> I believe that this is not enough a reason to keep WSL1 tests, and that,
> when we will have moved to GitHub actions, we should be able to have Mac
> tests instead, which would bring better value.
>
> I therefore think we should remove the WSL runners already now.
>
> Any thoughts?
>
> Cheers,
> Ben

Reply via email to