ci: isolate corrupt-cache reproduction from default test shards

The ``corrupt_cache_repro`` end-to-end test drives live
conda-forge in two independent solves (corrupt-cache vs.
clean-control) and runs ~2-5 minutes per parametrization. Putting
it in the default unit-test collection would dominate every
contributor's local pytest run and inflate every PR's CI wall
clock for negligible additional confidence on most changes.

Three pieces of CI plumbing:

1. ``pyproject.toml`` declares the ``corrupt_cache_repro`` marker
   and adds ``-m "not corrupt_cache_repro"`` to the default
   ``addopts``. ``pytest`` invocations from contributors,
   editors, and the unit-test CI shards exclude the e2e by
   default. Local opt-in is ``pytest -m corrupt_cache_repro``;
   bypass-everything is ``pytest -m ""``.

2. ``.github/workflows/test.yml`` adds a dedicated
   ``run-corrupt-cache-repro`` step that runs once per
   ``(os, py-version)`` on shard 1 / linux-64 only -- the
   fixture is a linux-64 cache snapshot from the PR conda#862
   harness, so other platforms have nothing to test against.
   The step uses ``-o "addopts=-rsx --verbose"`` to clear the
   project-default exclusion and ``-m corrupt_cache_repro`` to
   opt back into just the e2e parametrizations.

3. The CI step targets ``tests/e2e/test_corrupt_repodata_repro.py``
   directly so a future test author who marks an unrelated test
   with ``corrupt_cache_repro`` doesn't accidentally end up in
   the heavy CI lane.

Behavior of the unit + component tests is unchanged.

Verification: the new shard appears in the workflow matrix; the
default shards now deselect the marker via `-m 'not
corrupt_cache_repro'`. Confirm the partition is total by checking
that both invocations together select every test:
`pytest --collect-only -q -m corrupt_cache_repro` plus the
inverse should sum to `pytest --collect-only -q`.

Author: claude

.github/workflows/test.yml

@@ -142,6 +142,11 @@ jobs:
           mkdir -p tests/durations
           set -x
           which pytest
+          # Default `addopts` in pyproject.toml deselects the heavy
+          # live-conda-forge corrupt-cache reproduction (see the
+          # `corrupt_cache_repro` marker). The split unit-test job
+          # honours that exclusion; the dedicated step below opts back
+          # in and runs the e2e parametrizations on shard 1 only.
           pytest \
             --cov=conda_lock --cov-branch --cov-report=xml --cov-report=term \
             --store-durations \
@@ -151,6 +156,23 @@ jobs:
             --splits="${{ matrix.pytest-split-group-size }}" \
             --group="${{ matrix.pytest-split-group-index }}"
 
+      - name: run-corrupt-cache-repro
+        # Live conda-forge driven; ~5 min each parametrization. Run
+        # once per (os, python-version) on the first shard to keep
+        # total wall-clock down without losing coverage. Linux only --
+        # the fixture is a linux-64 cache snapshot from the PR #862
+        # harness.
+        if: matrix.pytest-split-group-index == 1 && matrix.os == 'ubuntu-latest'
+        run: |
+          set -x
+          # `-o "addopts="` clears the project-default `-m "not
+          # corrupt_cache_repro"`; then `-m corrupt_cache_repro` selects
+          # only the e2e parametrizations.
+          pytest \
+            -o "addopts=-rsx --verbose" \
+            -m corrupt_cache_repro \
+            tests/e2e/test_corrupt_repodata_repro.py
+
       - uses: codecov/codecov-action@671740ac38dd9b0130fbe1cec585b89eea48d3de # v5.5.2
         with:
           token: ${{ secrets.CODECOV_TOKEN }}

pyproject.toml

@@ -142,10 +142,27 @@ addopts = [
     "--doctest-modules",
     "--numprocesses=auto",
     "--ignore-glob=**/vendor_poetry/**",
-    "--ignore-glob=**/_vendor/**"
+    "--ignore-glob=**/_vendor/**",
+    # Heavy live-conda-forge reproduction test for conda/conda-lock#896
+    # is excluded from the default unit-test collection. CI's e2e job
+    # opts back in with `-m "corrupt_cache_repro"` (overrides this `-m`).
+    # Local invocation: `pytest -m corrupt_cache_repro` to run only
+    # those tests, or `pytest -m ""` to bypass the filter entirely.
+    "-m",
+    "not corrupt_cache_repro",
 ]
 log_level = "DEBUG"
 testpaths = ["tests", "conda_lock"]
+markers = [
+    # Heavy conda-forge-driven reproduction test for conda/conda-lock#896.
+    # Each parametrization runs ~2-5 minutes (real `conda-lock lock`
+    # against a hybrid cache, then `conda-lock render --kind=explicit`)
+    # and depends on live conda-forge state in TWO independent solves
+    # (corrupt-cache vs clean-control). Excluded from default unit
+    # collection via the `-m "not corrupt_cache_repro"` in `addopts`
+    # above; CI runs this marker in a dedicated e2e step.
+    "corrupt_cache_repro: heavy conda-forge-driven reproduction test for #896",
+]
 
 [tool.vendoring]
 destination = "conda_lock/_vendor"