CI parity: test the head merged into main

A head tested alone keeps failing harnesses that main already fixed. Since 10-25 evening, 9 of 21 failed PR
parity jobs failed only on such fixes: server_knockback (heads without #2114, 7 times), mob_ai's oracle heap
(without #2270), save_roundtrip (without #2212). #2160 needed 4 parity runs this way.

The job now merges main's current tip into the head and selects harnesses against that main commit, so only
the pull request's own changes count. A branch that doesn't change the runner scripts or the selection gets
main's, which replaces the two steps that copied main's vanilla-data.sh and parity.py in by hand. A head that
conflicts with main is tested alone, with a warning. The job summary names the main commit, and the check is
still `parity / parity` on the head.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

.github/workflows/parity.yml

@@ -26,35 +26,46 @@ jobs:
     steps:
       - uses: actions/checkout@v4
         with:
-          # The commit the event names: a fork's later pushes need another run-parity label.
+          # The commit the event names: a fork's later pushes need another run-parity label. The next step merges
+          # main into it.
           ref: ${{ inputs.head }}
           fetch-depth: 0
           persist-credentials: false
           # Keeps .local/ (parity state, probe builds, oracle results) and target/ between runs.
           clean: false
+      - name: The head merged into main (what merging it would land)
+        id: merge
+        env:
+          BASE_SHA: ${{ inputs.base }}
+          HEAD_SHA: ${{ inputs.head }}
+        run: |
+          # Harness fixes and game fixes that reached main since the branch was made then apply here too: a head
+          # tested alone failed harnesses already fixed on main (server_knockback before #2114, mob_ai's heap before
+          # #2270, save_roundtrip before #2212). Selection then compares with this main, so it sees only the pull
+          # request's own changes. A head that conflicts with main is tested alone.
+          set -euo pipefail
+          git merge --abort 2>/dev/null || true
+          git fetch -q --no-tags origin +refs/heads/main:refs/remotes/origin/main
+          main=$(git rev-parse origin/main)
+          if git -c user.name=parity -c user.email=parity@localhost merge -q --no-edit --no-ff "$main" >/dev/null 2>&1; then
+            echo "base=$main" >> "$GITHUB_OUTPUT"
+            echo "Tested: head ${HEAD_SHA:0:9} merged into main ${main:0:9}" | tee -a "$GITHUB_STEP_SUMMARY"
+          else
+            git merge --abort
+            echo "base=$BASE_SHA" >> "$GITHUB_OUTPUT"
+            echo "::warning::head ${HEAD_SHA:0:9} conflicts with main ${main:0:9}: tested alone, without main's fixes"
+            echo "Tested: head ${HEAD_SHA:0:9} alone (it conflicts with main ${main:0:9})" | tee -a "$GITHUB_STEP_SUMMARY"
+          fi
       - uses: dtolnay/rust-toolchain@stable
       - name: Tools from the runner
         run: bash tools/ci/runner-env.sh
       - name: Vanilla data, and resources prepared by this preparer
-        env:
-          BASE_SHA: ${{ inputs.base }}
-        run: |
-          # main's script: one written before a fix to it left a real .local/horizon-client without sounds.json on
-          # the runner, and every later job there failed `subtitle`. A branch that changes the script runs its own.
-          set -euo pipefail
-          if git rev-parse --verify -q origin/main >/dev/null \
-            && git diff --quiet "$(git merge-base "$BASE_SHA" HEAD)" HEAD -- tools/ci/vanilla-data.sh; then
-            git show origin/main:tools/ci/vanilla-data.sh > "$RUNNER_TEMP/vanilla-data.sh"
-            bash "$RUNNER_TEMP/vanilla-data.sh"
-          else
-            bash tools/ci/vanilla-data.sh
-          fi
+        run: bash tools/ci/vanilla-data.sh
       - name: The prepared output is the one the preparer's OUTPUT_VERSION says
         env:
-          BASE_SHA: ${{ inputs.base }}
+          BASE_SHA: ${{ steps.merge.outputs.base }}
         run: |
-          # The job runs main's workflow on the head's files: a head made before OUTPUT_VERSION (#1947) has neither
-          # the tool nor the number, so there is nothing to check.
+          # A head made before OUTPUT_VERSION (#1947) that conflicts with main has neither the tool nor the number.
           if [ ! -f tools/prepare_output.py ] || ! grep -q OUTPUT_VERSION crates/horizon-prepare/src/lib.rs; then
             echo "This head's preparer has no OUTPUT_VERSION: nothing to check"
             exit 0
@@ -65,47 +76,8 @@ jobs:
       - name: The harnesses this change can affect
         env:
           PARITY_HOSTS: ''
-          BASE_SHA: ${{ inputs.base }}
-        run: |
-          # main's selection (tools/parity.py --changed --list with main's parity.py and parity_reach.py), so a
-          # branch made before a better selection landed gets it too; the head's own runner then runs those
-          # harnesses. A branch that changes the selection itself selects with its own code.
-          set -euo pipefail
-          own() { python3 tools/parity.py --changed "$BASE_SHA" --timeout 2400; }  # world_spawn takes ~1480 s on a busy runner
-          fork_point=$(git merge-base "$BASE_SHA" HEAD)
-          if ! git rev-parse --verify -q origin/main >/dev/null || ! git diff --quiet "$fork_point" HEAD -- tools/parity.py tools/parity_reach.py; then
-            own
-            exit
-          fi
-          had_reach=$(git ls-files tools/parity_reach.py)
-          git show origin/main:tools/parity.py > tools/parity.py
-          git show origin/main:tools/parity_reach.py > tools/parity_reach.py
-          status=0
-          python3 tools/parity.py --changed "$BASE_SHA" --list > .local/parity-selection.txt || status=$?
-          git checkout -- tools/parity.py
-          if [ -n "$had_reach" ]; then git checkout -- tools/parity_reach.py; else rm -f tools/parity_reach.py; fi
-          cat .local/parity-selection.txt
-          if [ "$status" != 0 ]; then
-            echo "main's selection failed: the head's own selection runs"
-            own
-            exit
-          fi
-          names=$(sed '$d' .local/parity-selection.txt | grep -E '^[A-Za-z0-9_#]+$' || true)
-          # The summary line says how many would run: a count that doesn't match the names read (a format change)
-          # must not let the job pass untested.
-          expected=$(tail -1 .local/parity-selection.txt | sed -nE 's/^([0-9]+) of [0-9]+ harnesses would run$/\1/p')
-          count=$(printf '%s' "$names" | grep -c . || true)
-          if [ -z "$expected" ] || [ "$expected" != "$count" ]; then
-            echo "main's selection listed $count names, its summary says '${expected:-?}': the head's own selection runs"
-            own
-            exit
-          fi
-          if [ "$count" = 0 ]; then
-            echo "Nothing to run."
-            exit 0
-          fi
-          # shellcheck disable=SC2086
-          python3 tools/parity.py $names --timeout 2400
+          BASE_SHA: ${{ steps.merge.outputs.base }}
+        run: python3 tools/parity.py --changed "$BASE_SHA" --timeout 2400  # world_spawn takes ~1480 s on a busy runner
       - name: Logs of the failed harnesses
         if: failure()
         uses: actions/upload-artifact@v4

docs/release-flow.md

@@ -32,6 +32,12 @@ copy is: the Mac mini, his Linux desktop and playhorizon (below). Their runners
 between runs (`clean: false`), so the probes build incrementally into `.local/parity-target` and earlier passes
 stand.
 
+The job tests the pull request's head merged into main's current tip, which is what merging it would land: harness
+and game fixes that reached main after the branch was made apply to it too, and an old branch runs main's runner
+scripts and selection unless it changes them. The base for the selection is that main commit, so only the pull
+request's own changes count. The job summary names the main commit. A head that conflicts with main is tested
+alone, with a warning. The check is still reported on the head.
+
 1. `tools/ci/runner-env.sh` puts the machine's Python 3.13 and the JDK `research-lock.json` pins on the PATH
    (Homebrew and `java_home` on the Mac; on Linux what `linux-runner.sh` pinned, else nixpkgs), and sccache when
    the machine has it.
@@ -50,9 +56,6 @@ stand.
    modules, Java oracles and every Rust source their probes compile) changed since the merge base, plus any
    harness nothing is recorded about yet, and runs those that have no pass for their current inputs. A change to
    `Cargo.lock` or `research-lock.json` affects every harness. Each harness has 25 minutes, the job 45.
-   The selection runs with main's `tools/parity.py` and `tools/parity_reach.py` (`--changed <base> --list`), so a
-   branch made before a better selection landed gets it too, and the head's own `parity.py` then runs the named
-   harnesses. A branch that changes either file selects with its own code.
 
 Failed harnesses' logs are uploaded as the `parity-logs` artifact.