Merge branch 'master' into 4056.provides

.circleci/config.yml

@@ -24,6 +24,11 @@ version: 2.1
 dockerhub-context-template: &DOCKERHUB_CONTEXT
   context: "dockerhub-auth"
 
+# Required environment for using the coveralls tool to upload partial coverage
+# reports and then finish the process.
+coveralls-environment: &COVERALLS_ENVIRONMENT
+  COVERALLS_REPO_TOKEN: "JPf16rLB7T2yjgATIxFzTsEgMdN1UNq6o"
+
 # Next is a Docker executor template that gets the credentials from the
 # environment and supplies them to the executor.
 dockerhub-auth-template: &DOCKERHUB_AUTH
@@ -112,6 +117,21 @@ workflows:
       - "another-locale":
           {}
 
+      - "windows-server-2022":
+          name: "Windows Server 2022, CPython <<matrix.pythonVersion>>"
+          matrix:
+            parameters:
+              # Run the job for a number of CPython versions.  These are the
+              # two versions installed on the version of the Windows VM image
+              # we specify (in the executor).  This is handy since it means we
+              # don't have to do any Python installation work.  We pin the
+              # Windows VM image so these shouldn't shuffle around beneath us
+              # but if we want to update that image or get different versions
+              # of Python, we probably have to do something here.
+              pythonVersion:
+                - "3.9"
+                - "3.11"
+
       - "integration":
           # Run even the slow integration tests here.  We need the `--` to
           # sneak past tox and get to pytest.
@@ -126,6 +146,15 @@ workflows:
       - "docs":
           {}
 
+      - "finish-coverage-report":
+          requires:
+            # Referencing the job by "alias" (as CircleCI calls the mapping
+            # key) instead of the value of its "name" property causes us to
+            # require every instance of the job from its matrix expansion.  So
+            # this requirement is enough to require every Windows Server 2022
+            # job.
+            - "windows-server-2022"
+
   images:
     <<: *IMAGES
 
@@ -133,6 +162,20 @@ workflows:
     when: "<< pipeline.parameters.build-images >>"
 
 jobs:
+  finish-coverage-report:
+    docker:
+      - <<: *DOCKERHUB_AUTH
+        image: "python:3-slim"
+
+    steps:
+      - run:
+          name: "Indicate completion to coveralls.io"
+          environment:
+            <<: *COVERALLS_ENVIRONMENT
+          command: |
+            pip install coveralls==3.3.1
+            python -m coveralls --finish
+
   codechecks:
     docker:
       - <<: *DOCKERHUB_AUTH
@@ -151,6 +194,161 @@ jobs:
           command: |
             ~/.local/bin/tox -e codechecks
 
+  windows-server-2022:
+    parameters:
+      pythonVersion:
+        description: >-
+          An argument to pass to the `py` launcher to choose a Python version.
+        type: "string"
+        default: ""
+
+    executor: "windows"
+    environment:
+      # Tweak Hypothesis to make its behavior more suitable for the CI
+      # environment.  This should improve reproducibility and lessen the
+      # effects of variable compute resources.
+      TAHOE_LAFS_HYPOTHESIS_PROFILE: "ci"
+
+      # Tell pip where its download cache lives.  This must agree with the
+      # "save_cache" step below or caching won't really work right.
+      PIP_CACHE_DIR: "pip-cache"
+
+      # And tell pip where it can find out cached wheelhouse for fast wheel
+      # installation, even for projects that don't distribute wheels.  This
+      # must also agree with the "save_cache" step below.
+      PIP_FIND_LINKS: "wheelhouse"
+
+    steps:
+      - "checkout"
+
+      # If possible, restore a pip download cache to save us from having to
+      # download all our Python dependencies from PyPI.
+      - "restore_cache":
+          keys:
+            # The download cache and/or the wheelhouse may contain Python
+            # version-specific binary packages so include the Python version
+            # in this key, as well as the canonical source of our
+            # dependencies.
+            - &CACHE_KEY "pip-packages-v1-<< parameters.pythonVersion >>-{{ checksum \"setup.py\" }}"
+
+      - "run":
+          name: "Fix $env:PATH"
+          command: |
+            # The Python this job is parameterized is not necessarily the one
+            # at the front of $env:PATH.  Modify $env:PATH so that it is so we
+            # can just say "python" in the rest of the steps.  Also get the
+            # related Scripts directory so tools from packages we install are
+            # also available.
+            $p = py -<<parameters.pythonVersion>> -c "import sys; print(sys.prefix)"
+            $q = py -<<parameters.pythonVersion>> -c "import sysconfig; print(sysconfig.get_path('scripts'))"
+
+            New-Item $Profile.CurrentUserAllHosts -Force
+            # $p gets "python" on PATH and $q gets tools from packages we
+            # install.  Note we carefully construct the string so that
+            # $env:PATH is not substituted now but $p and $q are.  ` is the
+            # PowerShell string escape character.
+            Add-Content -Path $Profile.CurrentUserAllHosts -Value "`$env:PATH = `"$p;$q;`$env:PATH`""
+
+      - "run":
+          name: "Display tool versions"
+          command: |
+            python misc/build_helpers/show-tool-versions.py
+
+      - "run":
+          # It's faster to install a wheel than a source package.  If we don't
+          # have a cached wheelhouse then build all of the wheels and dump
+          # them into a directory where they can become a cached wheelhouse.
+          # We would have built these wheels during installation anyway so it
+          # doesn't cost us anything extra and saves us effort next time.
+          name: "(Maybe) Build Wheels"
+          command: |
+            if ((Test-Path .\wheelhouse) -and (Test-Path .\wheelhouse\*)) {
+              echo "Found populated wheelhouse, skipping wheel building."
+            } else {
+              python -m pip install wheel
+              python -m pip wheel --wheel-dir $env:PIP_FIND_LINKS .[testenv] .[test]
+            }
+
+      - "save_cache":
+          paths:
+            # Make sure this agrees with PIP_CACHE_DIR in the environment.
+            - "pip-cache"
+            - "wheelhouse"
+          key: *CACHE_KEY
+
+      - "run":
+          name: "Install Dependencies"
+          environment:
+            # By this point we should no longer need an index.
+            PIP_NO_INDEX: "1"
+          command: |
+            python -m pip install .[testenv] .[test]
+
+      - "run":
+          name: "Run Unit Tests"
+          environment:
+            # Configure the results location for the subunitv2-file reporter
+            # from subunitreporter
+            SUBUNITREPORTER_OUTPUT_PATH: "test-results.subunit2"
+
+            # Try to get prompt output from the reporter to avoid no-output
+            # timeouts.
+            PYTHONUNBUFFERED: "1"
+
+          command: |
+            # Run the test suite under coverage measurement using the
+            # parameterized version of Python, writing subunitv2-format
+            # results to the file given in the environment.
+            python -b -m coverage run -m twisted.trial --reporter=subunitv2-file --rterrors allmydata
+
+      - "run":
+          name: "Upload Coverage"
+          environment:
+            <<: *COVERALLS_ENVIRONMENT
+            # Mark the data as just one piece of many because we have more
+            # than one instance of this job (two on Windows now, some on other
+            # platforms later) which collects and reports coverage.  This is
+            # necessary to cause Coveralls to merge multiple coverage results
+            # into a single report.  Note the merge only happens when we
+            # "finish" a particular build, as identified by its "build_num"
+            # (aka "service_number").
+            COVERALLS_PARALLEL: "true"
+          command: |
+            python -m pip install coveralls==3.3.1
+
+            # .coveragerc sets parallel = True so we don't have a `.coverage`
+            # file but a `.coverage.<unique stuff>` file (or maybe more than
+            # one, but probably not).  coveralls can't work with these so
+            # merge them before invoking it.
+            python -m coverage combine
+
+            # Now coveralls will be able to find the data, so have it do the
+            # upload.  Also, have it strip the system config-specific prefix
+            # from all of the source paths.
+            $prefix = python -c "import sysconfig; print(sysconfig.get_path('purelib'))"
+            python -m coveralls --basedir $prefix
+
+      - "run":
+          name: "Convert Result Log"
+          command: |
+            # subunit2junitxml exits with error if the result stream it is
+            # converting has test failures in it!  So this step might fail.
+            # Since the step in which we actually _ran_ the tests won't fail
+            # even if there are test failures, this is a good thing for now.
+            subunit2junitxml.exe --output-to=test-results.xml test-results.subunit2
+
+      - "store_test_results":
+          path: "test-results.xml"
+
+      - "store_artifacts":
+          path: "_trial_temp/test.log"
+
+      - "store_artifacts":
+          path: "eliot.log"
+
+      - "store_artifacts":
+          path: ".coverage"
+
   pyinstaller:
     docker:
       - <<: *DOCKERHUB_AUTH
@@ -527,6 +725,15 @@ jobs:
   #     PYTHON_VERSION: "2"
 
 executors:
+  windows:
+    # Choose a Windows environment that closest matches our testing
+    # requirements and goals.
+    # https://circleci.com/developer/orbs/orb/circleci/windows#executors-server-2022
+    machine:
+      image: "windows-server-2022-gui:2023.06.1"
+      shell: "powershell.exe -ExecutionPolicy Bypass"
+    resource_class: "windows.large"
+
   nix:
     docker:
       # Run in a highly Nix-capable environment.

.coveragerc

@@ -19,7 +19,7 @@ skip_covered = True
 source =
 # It looks like this in the checkout
     src/
-# It looks like this in the Windows build environment
+# It looks like this in the GitHub Actions Windows build environment
     D:/a/tahoe-lafs/tahoe-lafs/.tox/py*-coverage/Lib/site-packages/
 # Although sometimes it looks like this instead.  Also it looks like this on macOS.
     .tox/py*-coverage/lib/python*/site-packages/

.github/workflows/ci.yml

@@ -44,13 +44,6 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os:
-          - windows-latest
-        python-version:
-          - "3.8"
-          - "3.9"
-          - "3.10"
-          - "3.11"
         include:
           # On macOS don't bother with 3.8, just to get faster builds.
           - os: macos-12

setup.py

@@ -413,7 +413,7 @@ setup(name="tahoe-lafs", # also set in __init__.py
               "pip==22.0.3",
               "wheel==0.37.1",
               "setuptools==60.9.1",
-              "subunitreporter==22.2.0",
+              "subunitreporter==23.8.0",
               "python-subunit==1.4.2",
               "junitxml==0.7",
               "coverage==7.2.5",

src/allmydata/test/cli/test_grid_manager.py

@@ -23,6 +23,9 @@ import click.testing
 from ..common_util import (
     run_cli,
 )
+from ..common import (
+    superuser,
+)
 from twisted.internet.defer import (
     inlineCallbacks,
 )
@@ -34,7 +37,6 @@ from twisted.python.runtime import (
 )
 from allmydata.util import jsonbytes as json
 
-
 class GridManagerCommandLine(TestCase):
     """
     Test the mechanics of the `grid-manager` command
@@ -223,7 +225,7 @@ class GridManagerCommandLine(TestCase):
             )
 
     @skipIf(platform.isWindows(), "We don't know how to set permissions on Windows.")
-    @skipIf(os.getuid() == 0, "cannot test as superuser with all permissions")
+    @skipIf(superuser, "cannot test as superuser with all permissions")
     def test_sign_bad_perms(self):
         """
         Error reported if we can't create certificate file

src/allmydata/test/common.py

@@ -117,6 +117,10 @@ from subprocess import (
     PIPE,
 )
 
+# Is the process running as an OS user with elevated privileges (ie, root)?
+# We only know how to determine this for POSIX systems.
+superuser = getattr(os, "getuid", lambda: -1)() == 0
+
 EMPTY_CLIENT_CONFIG = config_from_string(
     "/dev/null",
     "tub.port",

src/allmydata/test/test_client.py

@@ -77,6 +77,7 @@ from allmydata.scripts.common import (
 from foolscap.api import flushEventualQueue
 import allmydata.test.common_util as testutil
 from .common import (
+    superuser,
     EMPTY_CLIENT_CONFIG,
     SyncTestCase,
     AsyncBrokenTestCase,
@@ -151,7 +152,7 @@ class Basic(testutil.ReallyEqualMixin, unittest.TestCase):
     # EnvironmentError when reading a file that really exists), on
     # windows, please fix this
     @skipIf(platform.isWindows(), "We don't know how to set permissions on Windows.")
-    @skipIf(os.getuid() == 0, "cannot test as superuser with all permissions")
+    @skipIf(superuser, "cannot test as superuser with all permissions")
     def test_unreadable_config(self):
         basedir = "test_client.Basic.test_unreadable_config"
         os.mkdir(basedir)

src/allmydata/test/test_node.py

@@ -62,6 +62,7 @@ from .common import (
     ConstantAddresses,
     SameProcessStreamEndpointAssigner,
     UseNode,
+    superuser,
 )
 
 def port_numbers():
@@ -325,7 +326,7 @@ class TestCase(testutil.SignalMixin, unittest.TestCase):
         self.assertEqual(config.items("nosuch", default), default)
 
     @skipIf(platform.isWindows(), "We don't know how to set permissions on Windows.")
-    @skipIf(os.getuid() == 0, "cannot test as superuser with all permissions")
+    @skipIf(superuser, "cannot test as superuser with all permissions")
     def test_private_config_unreadable(self):
         """
         Asking for inaccessible private config is an error
@@ -341,7 +342,7 @@ class TestCase(testutil.SignalMixin, unittest.TestCase):
             config.get_or_create_private_config("foo")
 
     @skipIf(platform.isWindows(), "We don't know how to set permissions on Windows.")
-    @skipIf(os.getuid() == 0, "cannot test as superuser with all permissions")
+    @skipIf(superuser, "cannot test as superuser with all permissions")
     def test_private_config_unreadable_preexisting(self):
         """
         error if reading private config data fails
@@ -398,7 +399,7 @@ class TestCase(testutil.SignalMixin, unittest.TestCase):
         self.assertEqual(len(counter), 1) # don't call unless necessary
         self.assertEqual(value, "newer")
 
-    @skipIf(os.getuid() == 0, "cannot test as superuser with all permissions")
+    @skipIf(superuser, "cannot test as superuser with all permissions")
     def test_write_config_unwritable_file(self):
         """
         Existing behavior merely logs any errors upon writing