ci(update-docs): serialize concurrent runs that publish to MeshInspector.github.io

[Fedr]

Summary

Add a job-level concurrency: block to the update-documentation job in .github/workflows/update-docs.yml. GitHub will now serialize all runs that publish to the MeshInspector.github.io repo through a shared group, eliminating the git push race that produces:

! [rejected]          master -> master (fetch first)
error: failed to push some refs to 'https://github.com/MeshInspector/MeshInspector.github.io'

Why

When two CI runs (e.g. two PRs merged within minutes of each other, or a push and a schedule overlapping) finish near-simultaneously, the update-docs.yml job in each:

1. Checks out MeshInspector.github.io@master at slightly different points in time.
2. Regenerates docs locally.
3. Commits "Auto update docs".
4. Tries git push.

The first push wins; the second sees the ref has moved and is rejected, failing the entire update-dev-documentation job. Observed today in run 24936234953 (job update-documentation, scheduled run on master).

Change

 jobs:
   update-documentation:
+    concurrency:
+      group: meshinspector-github-io-publish
+      cancel-in-progress: false
     runs-on: ubuntu-latest

group: meshinspector-github-io-publish — any string; same string = same concurrency group = serialized.

cancel-in-progress: false — queue, don't cancel. Every commit's docs eventually publish; the second run waits for the first one's git push to land, then runs against the already-updated remote master and pushes its own commit on top. Trade-off vs. cancel-in-progress: true:

Setting	Behavior on overlapping runs	Throughput
false (this PR)	run B waits for run A, then publishes its own docs	every commit's docs make it eventually
true	run B cancels run A in flight; only B's docs publish	wastes A's Doxygen + artifact-download work

For a docs-publishing pipeline where each commit's docs are individually meaningful, false is the right default. We can flip to true later if we decide we only ever care about the latest doc snapshot.

Coverage

The concurrency block lives in the called workflow (update-docs.yml), so it protects both callers:

• build-test-distribute.yml → update-dev-documentation job → calls update-docs.yml (the one that hit the race today)
• update-docs-manual.yml → calls update-docs.yml (manual workflow_dispatch)

A scheduled run, a push-to-master run, and a manual run all serialize through the same group.

What this does not fix

Concurrency serializes CI-driven publishes only. If a human directly pushes to MeshInspector.github.io@master while a CI publish is in flight, the CI publish can still race against the human push. If that ever becomes a concern, a follow-up PR can wrap the push step in a pull --rebase + retry loop.

🤖 Generated with Claude Code