Re: [PATCH] ci: avoid unnecessary builds

From: Taylor Blau <me@ttaylorr.com>
Date: Nov 4, 2022, 02:23 UTC
Message-ID: <Y2R3vJf1A2KOZwA7@nand.local>
In-Reply-To: <221104.86bkpnzdqi.gmgdl@evledraar.gmail.com>

On Fri, Nov 04, 2022 at 02:46:23AM +0100, Ævar Arnfjörð Bjarmason wrote:

Show 7 quoted lines

> > This is particularly problematic in the case of Pull Requests where a
> > single contributor can easily (inadvertently) prevent timely builds for
> > other contributors.
>
> The "timely" being an issue in git/git and/or gitgitgadget where CI time
> is a shared resource, but not in a <someuser>/git running CI just for
> <someuser>?

Yup, agreed.

Show 7 quoted lines

> > To help with this situation, let's use the `concurrency` feature of
> > GitHub workflows, essentially canceling GitHub workflow runs that are
> > obsoleted by more recent runs:
> > https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#concurrency
>
> In my own fork I very much use this concurrency not-cancel-in-progress
> intentionally.

Interesting. I noted basically the opposite in my earlier reply[1] to Johannes, where the behavior I want is that newer pushes of the same topic supersede older ones that are currently in progress.

But I think you make a compelling point (which doesn't match my own workflow, but I can see the utility of nonetheless).

I was thinking that we could rely on something similar to the ci-config ref stuff from back in e76eec35540 (ci: allow per-branch config for GitHub Actions, 2020-05-07), but it looks like it'll be a little trickier than that, maybe impossible.

We need to know about the state of the ci-config branch before we set the concurrency bits. So I think you *could* do something like:

--- >8 ---

diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
index 4fdf4d3552..f1ca364f96 100644
--- a/.github/workflows/main.yml
+++ b/.github/workflows/main.yml
@@ -2,11 +2,6 @@ name: CI

 on: [push, pull_request]

-# Avoid unnecessary builds
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
 env:
   DEVELOPER: 1

@@ -39,7 +34,14 @@ jobs:
           then
             enabled=no
           fi
+          skip_concurrent=yes
+          if test -x config-repo/ci/config/skip-concurrent &&
+             ! config-repo/ci/config/skip-concurrent '${{ github.ref }}'
+          then
+            skip_concurrent=no
+          fi
           echo "::set-output name=enabled::$enabled"
+          echo "::set-output name=skip_concurrent::$skip_concurrent"
       - name: skip if the commit or tree was already tested
         id: skip-if-redundant
         uses: actions/github-script@v3
@@ -86,6 +88,9 @@ jobs:
     name: win build
     needs: ci-config
     if: needs.ci-config.outputs.enabled == 'yes'
+    concurrency:
+      group: ${{ github.workflow }}-${{ github.ref }}
+      cancel-in-progress: needs.ci-config.outputs.skip_concurrent = 'yes'
     runs-on: windows-latest
     steps:
     - uses: actions/checkout@v2
--- 8< ---

...and similar "concurrency" blocks in each of the other jobs to define
the settings at the job level instead of at the workflow level.

So, it's doable, but a little gross. At the very least, it would satisfy
Ævar's workflow requirements, too, since he could write a script that
exits with non-zero status to avoid the new behavior.

Thanks,
Taylor

[1]: https://lore.kernel.org/git/Y2R0YrQzKaUZzaPB@nand.local/