chore: use coverage run instead of pytest --cov (#3847)

* chore: improve code coverage reporting

* update after_n_builds

* fix: use `coverage run` instead of `pytest --cov` for accurate import-time coverage

The pytest11 entry point (`zarr = "zarr.testing"`) imports 76 zarr modules
during plugin loading, before pytest-cov starts measuring. This caused all
module-level code (class definitions, imports, decorators) to appear uncovered,
dropping reported coverage well below actual test coverage.

`coverage run -m pytest` starts tracing before any imports, resolving the issue.

* Revert changes

* Revert change

Author: maxrjones

.github/workflows/gpu_test.yml

@@ -72,4 +72,5 @@ jobs:
       uses: codecov/codecov-action@13ce06bfc6bbe3ecf90edbbf1bc32fe5978ca1d3  # v5.3.1
       with:
         token: ${{ secrets.CODECOV_TOKEN }}
+        flags: gpu
         verbose: true # optional (default = false)

.github/workflows/hypothesis.yaml

@@ -81,6 +81,7 @@ jobs:
       uses: codecov/codecov-action@v5
       with:
         token: ${{ secrets.CODECOV_TOKEN }}
+        flags: tests
         verbose: true # optional (default = false)
 
     - name: Generate and publish the report

.github/workflows/test.yml

@@ -71,6 +71,7 @@ jobs:
       uses: codecov/codecov-action@v5
       with:
         token: ${{ secrets.CODECOV_TOKEN }}
+        flags: tests
         verbose: true # optional (default = false)
 
   test-upstream-and-min-deps:
@@ -110,6 +111,7 @@ jobs:
       uses: codecov/codecov-action@v5
       with:
         token: ${{ secrets.CODECOV_TOKEN }}
+        flags: tests
         verbose: true # optional (default = false)
 
   doctests:

codecov.yml

@@ -8,9 +8,21 @@ coverage:
       default:
         target: auto
         threshold: 0.1
+        flags:
+          - tests
+flags:
+  tests:
+    paths:
+      - src/
+    carryforward: true
+  gpu:
+    paths:
+      - src/
+    carryforward: true
 codecov:
   notify:
-    after_n_builds: 10 # Wait for all 10 reports before updating the status
+    # 6 = test.yml: 3 (optional+ubuntu) + 2 (upstream + min_deps), hypothesis: 1
+    after_n_builds: 6
     wait_for_ci: yes
 comment:
   layout: "diff, files"

pyproject.toml

@@ -155,10 +155,6 @@ hooks.vcs.version-file = "src/zarr/_version.py"
 dependency-groups = ["test"]
 
 [tool.hatch.envs.test.env-vars]
-# Required to test with a pytest plugin; see https://pytest-cov.readthedocs.io/en/latest/plugins.html
-COV_CORE_SOURCE = "src"
-COV_CORE_CONFIG = ".coveragerc"
-COV_CORE_DATAFILE = ".coverage.eager"
 
 [[tool.hatch.envs.test.matrix]]
 python = ["3.12", "3.13", "3.14"]
@@ -175,13 +171,23 @@ matrix.deps.dependency-groups = [
 ]
 
 [tool.hatch.envs.test.scripts]
-run-coverage = "pytest --cov-config=pyproject.toml --cov=src --cov-append --cov-report xml --junitxml=junit.xml -o junit_family=legacy"
-run-coverage-html = "pytest --cov-config=pyproject.toml --cov=src --cov-append --cov-report html"
-run = "run-coverage --no-cov --ignore tests/benchmarks"
+run-coverage = [
+    "coverage run --source=src -m pytest --junitxml=junit.xml -o junit_family=legacy {args:}",
+    "coverage xml",
+]
+run-coverage-html = [
+    "coverage run --source=src -m pytest {args:}",
+    "coverage html",
+]
+run = "pytest --ignore tests/benchmarks"
 run-verbose = "run-coverage --verbose"
 run-mypy = "mypy src"
-run-hypothesis = "run-coverage -nauto --run-slow-hypothesis tests/test_properties.py tests/test_store/test_stateful*"
+run-hypothesis = [
+    "coverage run --source=src -m pytest -nauto --run-slow-hypothesis tests/test_properties.py tests/test_store/test_stateful* {args:}",
+    "coverage xml",
+]
 run-benchmark = "pytest --benchmark-enable tests/benchmarks"
+serve-coverage-html = "python -m http.server -d htmlcov 8000"
 list-env = "pip list"
 
 [tool.hatch.envs.gputest]
@@ -195,8 +201,11 @@ features = ["gpu"]
 python = ["3.12", "3.13"]
 
 [tool.hatch.envs.gputest.scripts]
-run-coverage = "pytest -m gpu --cov-config=pyproject.toml --cov=src --cov-report xml --junitxml=junit.xml -o junit_family=legacy --ignore tests/benchmarks"
-run = "run-coverage --no-cov"
+run-coverage = [
+    "coverage run --source=src -m pytest -m gpu --junitxml=junit.xml -o junit_family=legacy --ignore tests/benchmarks {args:}",
+    "coverage xml",
+]
+run = "pytest -m gpu --ignore tests/benchmarks"
 
 [tool.hatch.envs.upstream]
 template = 'test'