Nikolay Edigaryev

> I do like @nholloh's `TART_DOCKER_AUTH_CONFIG` name more though. Just `TART_DOCKER_CONFIG` seems not clear enough, Tart only uses this config for authentication. The `AUTH` part here is a bit misleading...

> Since this change is a proposal in the context of CI, specifically due to the issues we're facing on GitLab, I see value in following GitLab's naming scheme. This...

@nholloh that sounds like good trade-off to me, especially since we'd still need to change the GitLab Tart Executor to support this because GitLab Runner prepends `CUSTOM_ENV_` to CI/CD environment...

> If, however, the `TART_DOCKER_CONFIG` env var was set explicitly, I would expect tart to fail with an error and tell me the file was not found. Sounds reasonable to...

> Is there a call to action here? Yes, it would be nice to see native IOMMUFD support in Cloud Hypervisor.

You can use `--dirty`. Also see `--artifacts-dir`.

> My question is: what would be the easiest/recommended way to also make this work on Cirrus CI (the cloud)? I read [the docs for `artifacts` instruction](https://cirrus-ci.org/guide/writing-tasks/#artifacts-instruction), but didn't find...

> @edigaryev what do you think about a request header. Since the payload is a GraphQL query response and I don't think we can add `timestamp` there. 🤔 That would...

> Since the payload is a GraphQL query response and I don't think we can add `timestamp` there. 🤔 On a side note, it seems that the `action` and `old_status`...

Also, perhaps we should make the `audit_event` to conform to this format, or vice versa, because currently it differs, see screenshot below. Note that the `body` is not a nested...