fix: integration test failing on Windows CI

.gitignore

@@ -38,6 +38,10 @@ gurobi.log
 # earth-osm cache
 data/earth-osm/
 
+# test_snakemake_integration_testing's working directory; normally cleaned
+# up automatically, this is only a backstop if that's ever interrupted
+.integration-*/
+
 # Personal overrides (e.g. a custom Overpass API endpoint/token), passed via
 # --configfile config/config.local.yaml; never committed.
 config/config.local.yaml

tests/integration_test.py

@@ -10,6 +10,7 @@
 import shutil
 import subprocess
 import sys
+import tempfile
 from pathlib import Path
 
 import pytest
@@ -46,71 +47,85 @@ def test_standard_file_existance(module_path, file):
     assert Path(module_path / file).exists()
 
 
-def test_snakemake_integration_testing(module_path, tmp_path):
+def test_snakemake_integration_testing(module_path):
     """Run retrieval from scratch in its Conda environment and validate outputs."""
     log_dir = module_path / "tests/integration/logs"
     log_dir.mkdir(parents=True, exist_ok=True)
     run_log = log_dir / "integration.log"
     conda_prefix = module_path / ".snakemake/conda"
-    try:
-        with run_log.open("w", encoding="utf-8") as log:
-            result = subprocess.run(
-                [
-                    sys.executable,
-                    "-m",
-                    "snakemake",
-                    "--snakefile",
-                    str(module_path / "tests/integration/Snakefile"),
-                    "--directory",
-                    str(tmp_path),
-                    "--use-conda",
-                    "--conda-prefix",
-                    str(conda_prefix),
-                    "--cores",
-                    "1",
-                    "--show-failed-logs",
-                ],
-                stdout=log,
-                stderr=subprocess.STDOUT,
-                timeout=600,
-                check=False,
-                env={**os.environ, "XDG_CACHE_HOME": str(tmp_path / "cache")},
+
+    # A directory under module_path, not pytest's tmp_path: on Windows CI,
+    # tmp_path can land on a different drive than the checkout, and
+    # Snakemake's relative-path handling then fails with "ValueError: path
+    # is on mount 'D:', start on mount 'C:'". Keeping the working directory
+    # on the same drive as the Snakefile/conda prefix avoids that. Every
+    # assertion that reads from workdir has to stay inside this block -
+    # TemporaryDirectory deletes it on exit.
+    with tempfile.TemporaryDirectory(
+        dir=module_path, prefix=".integration-", ignore_cleanup_errors=True
+    ) as temporary_directory:
+        workdir = Path(temporary_directory)
+        try:
+            with run_log.open("w", encoding="utf-8") as log:
+                result = subprocess.run(
+                    [
+                        sys.executable,
+                        "-m",
+                        "snakemake",
+                        "--snakefile",
+                        str(module_path / "tests/integration/Snakefile"),
+                        "--directory",
+                        str(workdir),
+                        "--use-conda",
+                        "--conda-prefix",
+                        str(conda_prefix),
+                        "--cores",
+                        "1",
+                        "--show-failed-logs",
+                    ],
+                    stdout=log,
+                    stderr=subprocess.STDOUT,
+                    timeout=600,
+                    check=False,
+                    env={**os.environ, "XDG_CACHE_HOME": str(workdir / "cache")},
+                )
+        finally:
+            if (workdir / "logs").exists():
+                shutil.copytree(workdir / "logs", log_dir, dirs_exist_ok=True)
+
+        assert result.returncode == 0, run_log.read_text(encoding="utf-8")
+        output_dir = workdir / "resources/grid-builder/osm/retrieve"
+        for feature in ("substations_way", "lines_way"):
+            payload = json.loads(
+                (output_dir / f"benin_{feature}.json").read_text(encoding="utf-8")
             )
-    finally:
-        if (tmp_path / "logs").exists():
-            shutil.copytree(tmp_path / "logs", log_dir, dirs_exist_ok=True)
-
-    assert result.returncode == 0, run_log.read_text(encoding="utf-8")
-    output_dir = tmp_path / "resources/grid-builder/osm/retrieve"
-    for feature in ("substations_way", "lines_way"):
-        payload = json.loads(
-            (output_dir / f"benin_{feature}.json").read_text(encoding="utf-8")
-        )
-        elements = payload["elements"]
-        assert elements, f"No {feature} records retrieved"
-        assert all(item["geometry"] for item in elements)
-
-    build_dir = tmp_path / "resources/grid-builder/osm/build"
-    for component in ("buses", "lines", "transformers"):
-        with (build_dir / "csv" / f"{component}.csv").open(encoding="utf-8") as file:
-            rows = list(csv.DictReader(file))
-        assert rows or component == "transformers"
-
-    with (build_dir / "csv" / "lines.csv").open(encoding="utf-8") as file:
-        built_lines = list(csv.DictReader(file))
-    assert all(row["bus0"] != row["bus1"] for row in built_lines)
-    assert all(float(row["voltage_kv"]) >= 220 for row in built_lines)
-
-    for geojson_name in ("stations_polygon", "buses_polygon"):
-        polygons = json.loads(
-            (build_dir / "geojson" / f"{geojson_name}.geojson").read_text(
+            elements = payload["elements"]
+            assert elements, f"No {feature} records retrieved"
+            assert all(item["geometry"] for item in elements)
+
+        build_dir = workdir / "resources/grid-builder/osm/build"
+        for component in ("buses", "lines", "transformers"):
+            with (build_dir / "csv" / f"{component}.csv").open(
                 encoding="utf-8"
+            ) as file:
+                rows = list(csv.DictReader(file))
+            assert rows or component == "transformers"
+
+        with (build_dir / "csv" / "lines.csv").open(encoding="utf-8") as file:
+            built_lines = list(csv.DictReader(file))
+        assert all(row["bus0"] != row["bus1"] for row in built_lines)
+        assert all(float(row["voltage_kv"]) >= 220 for row in built_lines)
+
+        for geojson_name in ("stations_polygon", "buses_polygon"):
+            polygons = json.loads(
+                (build_dir / "geojson" / f"{geojson_name}.geojson").read_text(
+                    encoding="utf-8"
+                )
             )
-        )
-        assert polygons["type"] == "FeatureCollection"
-
-    country_log = (tmp_path / "logs/grid-builder/retrieve_osm_pbf/benin.log").read_text(
-        encoding="utf-8"
-    )
-    assert "Wrote" in country_log
-    assert str(conda_prefix) in country_log, "Retrieval must use its Conda Python"
+            assert polygons["type"] == "FeatureCollection"
+
+        country_log = (
+            workdir / "logs/grid-builder/retrieve_osm_pbf/benin.log"
+        ).read_text(encoding="utf-8")
+        assert "Wrote" in country_log
+        assert str(conda_prefix) in country_log, "Retrieval must use its Conda Python"