Re: Apply Patchseries Script

[Thomas Monjalon <thomas@monjalon.net>]

27/09/2023 18:31, Patrick Robb:
> Hello all,
> 
> The Community Lab is reviewing and rewriting some parts of our (currently
> internal) apply patchseries script. The reasons are:
> 
> 1. We want to remove any dependency the apply script has on our internal
> infrastructure, so that it can be upstreamed and utilized by others in the
> community.

Good news.

> 2. We want to add in new features like “depends-on patch” applying (like
> ovsrobot is doing currently)

Yes would be fantastic.

> 3. Some DPDK project processes have changed (like moving next branches from
> the main repo to being their own distinct repos). We have added on
> workarounds along the way to account for this, but an overall rework is now
> in order to clean up our process.

I don't think it changed.

> Before we do the work and attempt to upstream the script, I want to verify
> with the community that our current assumptions regarding the apply
> patchseries process are appropriate and should not be tweaked. Assumptions:
> 
> 1. There are two inputs, A. The pw series url and B. The branch output of
> pw_maintainers_cli.py

So it is only 1 input, because B can be deduced from A.

> 2. Do not apply and run if the series is an RFC series

Not sure about this requirement.
What is the problem in running tests on RFC?

> 3. Always check out to the current head of tree when applying a patch,
> regardless of whether the tree state has changed between patch submission
> and patch application in CI.

I don't think it is reasonable to look for the exact tree state
of patch submission, so yes I agree to use the head of the tree.
If it becomes quickly non applicable, then the author needs to update.
It does not happen frequently.

> 4. If the cover letter contains “depends-on,” extract the dependency series
> id(s), apply those, then attempt to apply the patch

Yes

> 5. If patch does not cleanly apply to the branch supplied by
> pw_maintainers_cli.py, attempt to apply on dpdk main. If this also fails,
> report an apply failure.

Yes

> 6. If apply is successful, attempt a sanity build, and report a build
> failure if that fails. If it succeeds, proceed with all CI testing.

Yes

> Note: The Community Lab does not currently use pw-client. If it is better
> for the CI community, we could stop maintaining a dedicated script for the
> apply process, try moving the pw-client, and direct our efforts at patching
> pw-client with the goal of adding support for features like depends-on. Are
> other labs using pw-client right now and do you recommend it?

In general I think it is a good idea to use common tools.
About adding depends-on support, it looks a great idea.
Other projects could use the same syntax then.
That's the same for the CI support in patchwork: we invented it in DPDK.