test: pin Rich colour and width so CLI-output tests stop failing outside CI

[saltas888]

Why

Tests that assert on CLI text compare against output whose colour and width Rich decides from the ambient environment. On a developer machine that exports FORCE_COLOR (or has a narrow terminal), Rich emits ANSI escapes and truncates its tables, so a large block of tests fails locally while staying green in CI — which trains contributors to explain away red test output.

This was reported independently by two contributors in the same week, in their PR descriptions: #1227 recorded 3 pre-existing failures in tests/unit/sdk/test_schema.py that "assert on plain strings while Rich emits ANSI colour codes in this environment", and #1233 recorded further unrelated-to-that-PR failures on stable.

Goal: make the suite's rendering environment deterministic, so CLI-output assertions give the same answer everywhere.

Non-goals: no production code changes, and no loosening of any assertion — the exact-output tests keep their value.

Closes INBOX-166

What changed

• tests/conftest.py — a pytest_configure hook pins the rendering environment for the whole suite: unset FORCE_COLOR, set NO_COLOR=1 and COLUMNS=200.
• tests/unit/sdk/test_schema.py — the three consoles the tests build themselves are now explicit: Console(file=StringIO(), width=1000, no_color=True, force_terminal=False), so they do not depend on the environment at all.
• tests/unit/test_render_env.py (new) — pins the pinning: that the hook ran before this module was imported, that the width survives any TERM, and that FORCE_COLOR is what would actually clamp it. A future edit that loosens the hook fails here with an explanation, instead of as a scatter of puzzling output-comparison failures across the ctl tests.
• tests/AGENTS.md — documents the invariant and the two traps below, so the next contributor does not rediscover them.

Implementation notes

Three details drove the design, all verified against rich 13.9.4 / click 8.3.3 rather than assumed:

1. It has to be pytest_configure, not a fixture. Rich snapshots no_color in Console.__init__, and many infrahub_sdk.ctl modules build a module-level Console() at import time — which happens while pytest collects the test modules, before any fixture can run. A session-scoped autouse fixture is already too late for those consoles.

2. FORCE_COLOR must be removed, not blanked. Rich treats any FORCE_COLOR value as proof it is writing to a terminal, empty string included (if force_color is not None: self._force_terminal = True).

3. TERM=dumb is deliberately not set. It looks like the obvious way to disable colour, but it is the trigger for Rich's dumb-terminal path, which pins the width to 80 and ignores COLUMNS — truncating the very wide tables the CLI-output fixtures record (e.g. the repository_list fixture is 139 columns).

   To be precise about the mechanism, since it was raised in review: TERM=dumb alone is harmless here, because is_dumb_terminal is is_terminal and TERM in ("dumb", "unknown") and captured test output is never a terminal. It is removing FORCE_COLOR (point 2) that keeps is_terminal False and so defuses the dumb path. Pinning TERM to a non-dumb value would therefore be a no-op for width while implying a knob that does nothing — whereas adopting TERM=dumb, as the card originally suggested, is what creates the failure if anything ever does force a terminal. test_render_env.py pins both halves of this.

One central hook covers all 24 CliRunner() call sites, so no per-test env plumbing was needed — passing env= to each runner would have been 24 edits that drift, and (per point 1) would not have fixed the module-level consoles anyway.

What stayed the same

No production code touched — the diff is tests/ plus a changelog fragment. No API, schema, dependency, or CI-workflow changes.

How to review

tests/conftest.py is the whole change; the rest follows from it. The comment block there records the three findings above.

How to test

The card's own acceptance gate, plus the environments that reproduce the original reports:

uv run pytest tests/unit/                                   # bare environment
FORCE_COLOR=1 COLUMNS=40 uv run pytest tests/unit/          # forced colour, narrow
NO_COLOR=1 uv run pytest tests/unit/                        # colour disabled
TERM=dumb uv run pytest tests/unit/                         # dumb terminal
FORCE_COLOR=3 TERM=xterm-256color COLUMNS=40 uv run pytest tests/unit/
FORCE_COLOR= TERM=dumb uv run pytest tests/unit/            # empty FORCE_COLOR + dumb
uv run ruff check . && uv run ruff format --check .

Before (on develop, with FORCE_COLOR=3 exported — the reported dev-machine case): 37 failed, 1781 passed, spread across tests/unit/ctl/test_marketplace_app.py (31), tests/unit/sdk/test_schema.py (3), tests/unit/ctl/test_repository_app.py (2) and tests/unit/ctl/test_task_app.py (1).

After: 1827 passed, 1 xfailed — identical under all six environments above. ruff check, ruff format --check, ty, mypy (158 files) and rumdl (129 files) are all clean; invoke lint reports only vale is not installed locally.

Two notes for the reviewer:

• The card also named test_gitrepo_init in tests/unit/sdk/test_repository.py. It does not fail on current develop in any of the environments tested, so nothing was changed for it.
• test_marketplace_app.py's 31 failures were not in the card's list but share the exact same root cause and are fixed by the same hook.

Impact & rollout

• Backward compatibility: test-only; no runtime behaviour changes.
• Config/env changes: the test process now sets NO_COLOR/COLUMNS and unsets FORCE_COLOR for itself. A side effect is that pytest's own report renders at 200 columns.
• Deployment notes: safe to merge; nothing to coordinate.

Checklist

• Tests added/updated
• Changelog entry added
• External docs updated (if user-facing or ops-facing change) — n/a, test-only
• Internal .md docs updated (tests/AGENTS.md)

---

Filed from Engineering Inbox card INBOX-166 by the platform-health agent. Opened for human review — the agent does not merge.

🤖 Generated with Claude Code

---

Summary by cubic

Pins Rich's colour and width for the test suite so CLI-output assertions behave the same everywhere, fixing 37 tests that failed on developer machines but passed in CI. Closes INBOX-166.

• tests/conftest.py now unsets FORCE_COLOR and sets NO_COLOR=1 and COLUMNS=200 from pytest_configure before any test module is imported.
• The three consoles in tests/unit/sdk/test_schema.py are constructed with no_color=True and force_terminal=False so they don't depend on the environment.
• tests/unit/test_render_env.py guards the pinning: the env is applied, the width survives any TERM, and FORCE_COLOR set with TERM=dumb is pinned as the failure mode the hook prevents, for every value including the empty string.

^{Written for commit caada89. Summary will update on new commits.}

[cloudflare-workers-and-pages]

Deploying infrahub-sdk-python with    Cloudflare Pages

Latest commit:	caada89
Status:	✅  Deploy successful!
Preview URL:	https://1ea6d260.infrahub-sdk-python.pages.dev
Branch Preview URL:	https://pha-inbox-166.infrahub-sdk-python.pages.dev

View logs

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

@@             Coverage Diff             @@
##           develop    #1332      +/-   ##
===========================================
- Coverage    85.54%   84.24%   -1.31%     
===========================================
  Files          148      147       -1     
  Lines        14271    13047    -1224     
  Branches      1953     1930      -23     
===========================================
- Hits         12208    10991    -1217     
+ Misses        1496     1493       -3     
+ Partials       567      563       -4     

Flag	Coverage Δ
integration-tests	39.10% <ø> (-4.86%)	⬇️
python-3.10	57.07% <ø> (-3.62%)	⬇️
python-3.11	57.06% <ø> (-3.65%)	⬇️
python-3.12	57.06% <ø> (-3.65%)	⬇️
python-3.13	57.07% <ø> (-3.64%)	⬇️
python-3.14	57.07% <ø> (-3.62%)	⬇️
python-filler-3.12	23.68% <ø> (+2.03%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 4 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[cubic-dev-ai]

All reported issues were addressed across 4 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[cubic-dev-ai]

All reported issues were addressed across 1 file (changes from recent commits).

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>