[infra] Preserve & replace nightly artifacts across reruns (#57)

* [infra] Preserve nightly artifacts across reruns

Defer aggregation and cleanup after matrix failures so selective job reruns can reuse successful jobs' staging artifacts.

* [infra] Replace nightly artifacts on reruns

Overwrite fixed-name artifacts so selective retries publish current outputs instead of colliding with stale results from earlier attempts.

Author: vikuznetsov-nvidia

.github/workflows/nightly.yml

@@ -272,6 +272,7 @@ jobs:
         with:
           path: output/dist/cuvslam-docs-${{ needs.check-changes.outputs.package_version }}.tar.gz
           archive: false
+          overwrite: true
           retention-days: 30
           if-no-files-found: error
 
@@ -294,6 +295,7 @@ jobs:
             output/eval/kpi_*.json
             output/eval/kpi_*.json.table
             output/eval/kpi_*.json.drift
+          overwrite: true
           retention-days: 1
           if-no-files-found: warn
 
@@ -303,6 +305,7 @@ jobs:
         with:
           name: ${{ env.EVAL_REPORT_ARTIFACT }}-${{ needs.check-changes.outputs.package_version }}-${{ matrix.platform }}-cuda${{ matrix.cuda }}-ubuntu${{ matrix.ubuntu }}
           path: output/eval/stats/**/report_*.pdf
+          overwrite: true
           retention-days: 1
           if-no-files-found: warn
 
@@ -315,6 +318,7 @@ jobs:
             output/cpp-test-results.xml
             output/python-test-output.txt
             output/test-summary.txt
+          overwrite: true
           retention-days: 1
           if-no-files-found: warn
 
@@ -324,6 +328,7 @@ jobs:
         with:
           path: output/dist/cuvslam-cpp-${{ needs.check-changes.outputs.package_version }}-${{ matrix.platform }}-cuda${{ matrix.cuda }}-ubuntu${{ matrix.ubuntu }}.tar.gz
           archive: false
+          overwrite: true
           retention-days: 30
           if-no-files-found: error
 
@@ -333,6 +338,7 @@ jobs:
         with:
           path: output/wheel/*.whl
           archive: false
+          overwrite: true
           retention-days: 30
           if-no-files-found: error
 
@@ -349,23 +355,29 @@ jobs:
       contents: read
     steps:
       - name: Download test result artifacts
-        if: needs.check-changes.outputs.should_build == 'true'
+        if: >-
+          needs.check-changes.outputs.should_build == 'true' &&
+          needs.build-and-test.result == 'success'
         continue-on-error: true
         uses: actions/download-artifact@v8
         with:
           pattern: ${{ env.TEST_RESULTS_ARTIFACT }}-${{ needs.check-changes.outputs.package_version }}-*
           path: artifacts/test-results
 
       - name: Download KPI staging artifacts
-        if: needs.check-changes.outputs.should_build == 'true'
+        if: >-
+          needs.check-changes.outputs.should_build == 'true' &&
+          needs.build-and-test.result == 'success'
         continue-on-error: true
         uses: actions/download-artifact@v8
         with:
           pattern: ${{ env.EVAL_KPIS_ARTIFACT }}-${{ needs.check-changes.outputs.package_version }}-*
           path: artifacts/kpis
 
       - name: Download evaluation report staging artifacts
-        if: needs.check-changes.outputs.should_build == 'true'
+        if: >-
+          needs.check-changes.outputs.should_build == 'true' &&
+          needs.build-and-test.result == 'success'
         continue-on-error: true
         uses: actions/download-artifact@v8
         with:
@@ -412,6 +424,17 @@ jobs:
             exit 0
           fi
 
+          if [ "${{ needs.build-and-test.result }}" != "success" ]; then
+            {
+              echo "Artifact consolidation is deferred until all matrix jobs succeed."
+              echo "Per-configuration staging artifacts are retained for selective job reruns."
+              echo ""
+            } >> "$REPORT"
+            echo "has_test_results=false" >> "$GITHUB_OUTPUT"
+            echo "has_reports=false" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+
           test_config_count=0
           TEST_ARTIFACT_PREFIX="${TEST_RESULTS_ARTIFACT}-${PACKAGE_VERSION}-"
           for testdir in artifacts/test-results/${TEST_ARTIFACT_PREFIX}*; do
@@ -517,6 +540,7 @@ jobs:
         with:
           name: test-results-${{ needs.check-changes.outputs.package_version }}
           path: consolidated/test-results/
+          overwrite: true
           retention-days: 30
           if-no-files-found: error
 
@@ -552,12 +576,13 @@ jobs:
         with:
           path: consolidated/cuvslam-evaluation-${{ needs.check-changes.outputs.package_version }}.tar.gz
           archive: false
+          overwrite: true
           retention-days: 30
           if-no-files-found: error
 
       - name: Delete staging artifacts
         id: finalize-artifacts
-        if: always()
+        if: always() && needs.build-and-test.result == 'success'
         uses: actions/github-script@v7
         env:
           PACKAGE_VERSION: ${{ needs.check-changes.outputs.package_version }}