On Wed, Dec 07 2022, Johannes Schindelin via GitGitGadget wrote:
> I waited for quite a while with submitting this because I did not want to
> interfere with patch series that are in flight, but it seems that this was
> unwise, as there is now a patch floating about (ab/ci-retire-set-output)
> that partially fulfills this here patch series' purpose. However, these
> patches are more complete, so I proposed to supersede that patch with this
> more comprehensive solution.
What does "more complete" here mean? Just that this series is doing more
things than stand-alone patches?
> This patch series is based on od/ci-use-checkout-v3-when-applicable.
>
> Johannes Schindelin (4):
> ci: use a newer `github-script` version
Is this the patch that needs or interacts with
od/ci-use-checkout-v3-when-applicable?
I don't see it in "What's Cooking", but "seen" currently has
"gitster/js/ci-set-output", which looks to be from
[1]
> ci: avoid deprecated `set-output` workflow command
This is byte-for-byte the same as the second hunk in my [2].
> ci: avoid using deprecated {up,down}load-artifacts Action
Most of this looks good & doesn't duplicate any existing patch, but why
is there a change in there that disables the uploading of failed test
directories for the "linux32" job?
> ci(l10n): avoid using the deprecated `set-output` workflow command
This does the same as the first hunk in my [2], you're just using two
"echo", I used a here-doc.
I think what would make it valuable to bundle these up is if doing so
would resolve some tricky interference between these patches.
But I don't see that that's the case, and we can e.g. start writing to
"$GITHUB_OUTPUT" independent of other changes.
1. https://lore.kernel.org/git/pull.1387.git.1667902408921.gitgitgadget@xxxxxxxxx/
2. https://lore.kernel.org/git/patch-v2-1.1-4e7db0db3be-20221207T014848Z-avarab@xxxxxxxxx/
