Verify the version in the package workflow, not in a unit test - #1647

test_get_version asserts that get_version() returns a version matching a hand-written pattern: release[.devN[+local]]. On speedkick that assertion fails on every pull request:

FAILED packages/server/server_tests/memmachine_server/common/api/test_version.py::test_get_version
AssertionError: Invalid server version: 0.3.9.post2.dev14+g2c69bb83e

main passes it today, and this PR is the same change as #1604 (merged on speedkick) applied here, so the two branches carry the same test file and the same package-workflow check, and so main does not start failing the moment the cause becomes reachable from it.

Cause

The annotated tag v0.3.9-post1 ("Post-release build from speedkick, after v0.3.9", 2026-08-31) sits on speedkick's ancestry and is the nearest tag for every commit since. setuptools-scm therefore renders each speedkick build as 0.3.9.post2.devN+g<sha>. The test's pattern never allowed a post-release segment, so it rejected a version setuptools-scm produces by design. main is unaffected only because that tag is not yet reachable from it — git describe on main gives v0.3.9-22-g..., on speedkick v0.3.9-post1-21-g.... Merging speedkick into main, or tagging any post-release on main, reproduces the failure here.

Why the test is deleted rather than patched

get_version() is a wrapper around importlib.metadata.version for two distribution names. Nothing in it produces a version shape; the shape comes from setuptools-scm, the repository's tags and the working tree. So the test could only ever fail for environmental reasons, and that is its whole history:

The function under test has not changed since #951 wrote it. The test also rejected a documented output: #951 specifies client_version as "not available" when memmachine-client is not installed, which is the case for a wheel install of the server (the server package does not depend on the client). It passed only because the uv workspace installs every member.

test_version_string stays; it tests the model's own string form.

Where the check goes instead

The Test Server Package workflow already builds the wheels and installs them. A new step runs the installed memmachine-server --version and requires the reported server version to equal the version in the wheel's filename. That ties the binary to the artifact it shipped in without asserting anything about the version's format, so tag conventions can change without touching it.

Verification

• The trimmed test file passes and is clean under ruff check and ruff format --check.
• The new workflow step, extracted verbatim from the YAML and run against wheels built from this tree and installed into a fresh virtualenv, passes — and reports client: not available, which the deleted test would have rejected.

Out of scope: the windows unit-test step runs under pwsh

The pytest workflow's test step has no shell: key, so on windows it runs under pwsh, which does not stop on the first command's non-zero exit and returns the status of the last command (the client suite). A server-suite failure on windows would not fail the job. On main the windows server run passes today (1830 passed, 4 skipped on #1584's run), so nothing is hidden yet; on speedkick two windows-only failures are, and forcing bash there would turn windows red until they are fixed. Left as is here for parity.

🤖 Generated with Claude Code