Count dists before uploading scientific wheels or publishing to PyPI

[radarhere]

Suggestions for python-pillow#9239

• Update actions/download-artifact to v5
• Count dists before uploading scientific wheels or publishing to PyPI. If we suddenly started producing more wheels than expected, maybe we should inspect that before sharing those wheels?
• At the moment, the printed number of dists and the number of dists checked by the condition are from two different ls operations. I suggest making that the same operation, so the user can be confident in what the output is saying.

[radarhere]

The iOS arm64 simulator failure is because your fork is using macos-15-arm64. The main repository is still on macos-14-arm64 for the moment, so I expect this to pass if merged.

I find this odd, as we are requesting macos-14.

[hugovk]

The push and pull_request triggers are slightly different.

• push just runs the commit as it appears in the branch.

• But pull_request simulates merging the commit into the base, and runs that.

Most of the time, it's the same result. But in cases like this, it fails. And we only have pull_request running here.

I'll update my main from upstream and restart this.

[hugovk]

And the failing run here has:

          - name: "iOS arm64 simulator"
            platform: ios
            os: macos-latest

• https://github.com/hugovk/Pillow/actions/runs/18423698980/workflow?pr=139#L101

Whereas upstream has:

          - name: "iOS arm64 simulator"
            platform: ios
            os: macos-14

https://github.com/python-pillow/Pillow/actions/runs/18423923094/workflow#L100

[hugovk]

Let's see if close/re-open triggers it.

[hugovk]

Close/re-open didn't do it: https://github.com/hugovk/Pillow/actions/runs/18427497312/workflow?pr=139#L101

And wow, the rebase on upstream/main didn't help either!

[hugovk]

OK, that wasn't a pretty Git session but we're there now!

[radarhere]

I don't know what's going on with this PR that is causing lint to fail, but I remain confident that the changes would be valid for python-pillow.

[hugovk]

I don't know what's going on with this PR that is causing lint to fail, but I remain confident that the changes would be valid for python-pillow.

Yeah, let's just merge and move things back there :)