Erik Sipsma

Are there any updates on the progress here? The only reason I care is that https://github.com/moby/buildkit/pull/4455 is accidentally dependent on it (see the PR description for why). I totally get...

One thing I was thinking about during merge+diff is that it would be nice if laziness was implemented entirely on the solver level rather than in the cache package. Basic...

@tonistiigi The thought process behind my suggestion is basically: 1. Laziness seems like a decision about scheduling the execution of work, which feels like it would fit better on the...

Been thinking about how to clean up lazy refs more. I agree that the fundamental problem is that we store refs that can't be loaded without external context. The general...

> related but from the distance. The CLI will eventually use the CLI helper, but for the moment, according to what we spoke with @sipsma and @aluzzardi we'll keep using...

> This one I don't understand: since the CLI relies on the Go SDK, isn't all this logic already handled in the Go SDK? And if so, why would we...

To clarify one thing, `binfmt` isn't just a package that needs to be installed, the `docker run --rm --privileged tonistiigi/binfmt:latest --install all` command actually results in the kernel being configured...

@grouville Took a look, the tests all pass on my macbook as is (starting from clean slate, no local images pulled or anything) and I found that if I decrease...

> we found an agreement on the future of this PR: make it a periodic check (once a day), and rename it as as automated-provisioning-test. SGTM! > Do you think...

> Right now macOS only (maybe Linux wouldn't hurt?) but someday Windows probably. SGTM, Linux we actually will be able to test w/ dagger-in-dagger once we have docker socket mount...