Fix locale-sensitive Windows tests and technical token formatting

[TheAngryPit]

Tracking issue: #784

Summary

This PR makes a small set of Windows test/formatting paths safe on localized Windows installations:

• Compare sandbox Documents, Desktop, and Downloads grants against OS-provided known-folder paths instead of English folder suffixes.
• Format compact technical token counts with InvariantCulture, so regional decimal separators do not change test-facing output.
• Serialize tests that mutate AppVersionInfo.TestOverride, including WindowsNodeClientTests, to avoid parallel shared-state races.
• Mark shared test projects explicitly so local/CI validation does not silently no-op.

This is PR1 in the Windows 25H2 / MXC validation tracking issue. It is independent from the MXC runtime enablement PR and can be reviewed separately.

Why

On a Portuguese Windows setup, known folders can resolve to localized or redirected locations such as Documentos and Ambiente de Trabalho. The previous tests assumed English path suffixes like Documents, Desktop, or Downloads, which makes otherwise correct behavior look broken on localized Windows.

The compact token count formatter also used current-culture numeric formatting. That can produce comma decimals such as 15,3K in locales where the technical/test contract expects stable dot-decimal output.

Changed Files

• src/OpenClaw.Tray.WinUI/Chat/ChatUsageFormatter.cs
• tests/Directory.Build.props
• tests/OpenClaw.Shared.Tests/AppVersionInfoTestCollection.cs
• tests/OpenClaw.Shared.Tests/AppVersionInfoTests.cs
• tests/OpenClaw.Shared.Tests/Mxc/MxcPolicyBuilderTests.cs
• tests/OpenClaw.Shared.Tests/WindowsNodeClientTests.cs

Validation

Current-head validation was repeated for commit 0555bfa52c0de1d6b88055260d24995d2c2b7782 on Windows Release Preview 25H2 build 26200.8728, Portuguese Windows locale:

• ./build.ps1: passed; all builds succeeded.
• dotnet test ./tests/OpenClaw.Shared.Tests/OpenClaw.Shared.Tests.csproj --no-restore: passed; 2262 passed, 29 skipped, 0 failed.
• dotnet test ./tests/OpenClaw.Tray.Tests/OpenClaw.Tray.Tests.csproj --no-restore: passed; 1088 passed, 0 skipped, 0 failed.
• openclaw-autoreview --mode branch --base upstream/main --engine codex --model gpt-5.5 --thinking high: clean; no accepted/actionable findings; overall patch is correct (0.86).

Validation note: this was run in a fresh worktree at the exact PR head. The first --no-restore test command hit the repository-documented first-run project.assets.json gotcha, so the test projects were restored and the full required validation sequence was repeated.

Notes

This PR intentionally does not change MXC availability, system.run, Gateway pairing, WSL Gateway setup, release versioning, build automation, validation helper scripts, or installed Companion behavior.

The MXC runtime and Gateway proof are split into #786 and #787. This PR remains only the locale/test baseline and intentionally does not claim MXC runtime behavior.

[clawsweeper]

Codex review: needs maintainer review before merge. Reviewed June 20, 2026, 4:08 PM ET / 20:08 UTC.

Summary
The PR makes compact chat token counts culture-invariant, updates localized Windows known-folder tests to compare OS-resolved paths, serializes AppVersionInfo-mutating tests, and marks test projects explicitly.

Reproducibility: yes. Source inspection shows current main uses culture-sensitive count formatting and English suffix assertions for localized Windows folders, and the contributor supplied current-head Portuguese Windows validation.

Review metrics: 2 noteworthy metrics.

• Changed surface: 6 files, +56/-8. The actual merge result is small and limited to one formatter file plus tests/test metadata.
• Required proof commands: 3 reported passed. The contributor reports the AGENTS-required build, Shared.Tests, and Tray.Tests all passing at the current PR head on the affected locale setup.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #784
Summary: This PR is the candidate fix for the locale/test baseline slice of the Windows 25H2/MXC validation tracker; the sibling PRs cover adjacent runtime-policy and Gateway-proof slices.

Members:

• canonical: Track Windows 25H2 build 26200.8728 MXC validation #784 - The tracker explicitly lists this PR as the first locale/test baseline slice and remains open for stack-level maintainer decisions.
• adjacent_distinct: Validate MXC SDK 0.7.0 on Windows build 26200.8728 #786 - That open follow-up depends on this baseline but changes MXC SDK/runtime policy rather than locale-safe tests and token formatting.
• adjacent_distinct: Add Gateway system.run MXC runtime E2E proof #787 - That open follow-up is stacked later for Gateway system.run MXC runtime proof and does not replace this baseline.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🦞 diamond lobster
Proof: 🦞 diamond lobster
Patch quality: 🦞 diamond lobster
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Risk before merge

• [P1] The open follow-up PRs Validate MXC SDK 0.7.0 on Windows build 26200.8728 #786 and Add Gateway system.run MXC runtime E2E proof #787 depend on this baseline and may need rebase or merge-order coordination after it lands.

Maintainer options:

1. Decide the mitigation before merge
   Land this narrowed locale/test baseline if maintainer review agrees, then rebase and review the MXC runtime-policy and Gateway-proof follow-up PRs separately.
2. Pause or close
   Do not merge this PR until maintainers decide whether the risk is worth taking.

Next step before merge

• No automated repair is needed; the remaining action is maintainer review and merge-order coordination with the linked follow-up PRs.

Security
Cleared: The diff touches formatter logic, tests, and test project metadata only; I found no concrete security or supply-chain concern.

Review details

Best possible solution:

Land this narrowed locale/test baseline if maintainer review agrees, then rebase and review the MXC runtime-policy and Gateway-proof follow-up PRs separately.

Do we have a high-confidence way to reproduce the issue?

Yes. Source inspection shows current main uses culture-sensitive count formatting and English suffix assertions for localized Windows folders, and the contributor supplied current-head Portuguese Windows validation.

Is this the best way to solve the issue?

Yes. The narrowed PR fixes the behavior at the right boundary by using invariant formatting, OS-resolved folder expectations, and test serialization without changing MXC runtime policy.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against c0514bd2d026.

Label changes

Label changes:

• add rating: 🦞 diamond lobster: Overall readiness is 🦞 diamond lobster; proof is 🦞 diamond lobster and patch quality is 🦞 diamond lobster.
• remove rating: 🐚 platinum hermit: Current PR rating is rating: 🦞 diamond lobster, so this older rating label is no longer current.

Label justifications:

• P2: This is a normal-priority Windows locale/test reliability and technical formatting fix with limited production blast radius.
• rating: 🦞 diamond lobster: Overall readiness is 🦞 diamond lobster; proof is 🦞 diamond lobster and patch quality is 🦞 diamond lobster.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (terminal): The PR body and latest comment provide current-head Windows 25H2 Portuguese-locale build/test/autoreview output with pass counts for the changed behavior.
• proof: sufficient: Contributor real behavior proof is sufficient. The PR body and latest comment provide current-head Windows 25H2 Portuguese-locale build/test/autoreview output with pass counts for the changed behavior.

Evidence reviewed

What I checked:

• Repository policy read: AGENTS.md was read in full; its required validation sequence and first-run test gotcha apply to the proof assessment for this PR. (AGENTS.md:1, c0514bd2d026)
• Current main formatter is culture-sensitive: Current main formats compact token counts with interpolated numeric format strings and value.ToString(), so decimal separators can follow the current culture. (src/OpenClaw.Tray.WinUI/Chat/ChatUsageFormatter.cs:90, c0514bd2d026)
• PR merge ref uses invariant formatting: The actual merge ref changes the formatter to CultureInfo.InvariantCulture for millions, thousands, and raw counts. (src/OpenClaw.Tray.WinUI/Chat/ChatUsageFormatter.cs:97, 5a3d6c02f49e)
• Current main tests assume English folder suffixes: Current main MXC policy tests assert Documents, Desktop, and Downloads by English suffix, matching the reported localized Windows brittleness. (tests/OpenClaw.Shared.Tests/Mxc/MxcPolicyBuilderTests.cs:115, c0514bd2d026)
• PR merge ref compares OS-resolved folder paths: The merge ref compares Documents, Desktop, and Downloads grants against Environment.GetFolderPath or the same Downloads known-folder API fallback used by production code. (tests/OpenClaw.Shared.Tests/Mxc/MxcPolicyBuilderTests.cs:115, 5a3d6c02f49e)
• Merge result is narrow: GitHub's merge ref changes only the six intended formatter/test files, 56 additions and 8 deletions, so stale branch drift is not deleting current-main-only files in the actual merge result. (5a3d6c02f49e)

Likely related people:

• the99missedcalls: git blame points the compact token formatter changed by this PR to the chat usage footer feature commit. (role: introduced formatter behavior; confidence: high; commits: ff38d8499c48; files: src/OpenClaw.Tray.WinUI/Chat/ChatUsageFormatter.cs, tests/OpenClaw.Tray.Tests/ChatUsageFormatterTests.cs)
• AlexAlves87: git blame/log point the current MXC policy builder/tests, AppVersionInfo tests, and shared test props to this area history. (role: shared-test and MXC policy test area contributor; confidence: high; commits: b175439fbfd4; files: src/OpenClaw.Shared/Mxc/MxcPolicyBuilder.cs, tests/OpenClaw.Shared.Tests/Mxc/MxcPolicyBuilderTests.cs, tests/OpenClaw.Shared.Tests/AppVersionInfoTests.cs)
• Ranjesh: git log shows recent gateway connection and pairing hardening in WindowsNodeClientTests, adjacent to the unsafe request-id assertion adjusted here. (role: recent WindowsNodeClient test contributor; confidence: medium; commits: ea36b12f9e4c; files: tests/OpenClaw.Shared.Tests/WindowsNodeClientTests.cs, src/OpenClaw.Shared/WindowsNodeClient.cs)

What the crustacean ranks mean

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

[TheAngryPit]

Status update after latest push (debafff):

• Fixed the origin/main zero-test/no-op risk by marking shared test projects explicitly with IsTestProject=true.
• Fixed the brittle unsafe request-id assertion by checking the full unsafe id (
  eq-1 && bad) instead of the incidental substring �ad.
• Local CI-equivalent proof: Shared.Tests executed 2262/passed 2262; Tray.Tests executed 1088/passed 1088.
• Autoreview rerun: clean, no accepted/actionable findings.
• Review threads: none unresolved.

@clawsweeper re-review

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

[TheAngryPit]

@clawsweeper re-review

Current-head validation was repeated for 0555bfa52c0de1d6b88055260d24995d2c2b7782.

Scope:

• Fresh detached worktree at the exact PR head.
• Windows Release Preview 25H2 build 26200.8728.
• Portuguese Windows locale.
• No code changes in this pass; this is proof refresh for the current head.

Validation:

• ./build.ps1 -> passed; all builds succeeded.
• dotnet test ./tests/OpenClaw.Shared.Tests/OpenClaw.Shared.Tests.csproj --no-restore -> passed; 2262 passed, 29 skipped, 0 failed.
• dotnet test ./tests/OpenClaw.Tray.Tests/OpenClaw.Tray.Tests.csproj --no-restore -> passed; 1088 passed, 0 skipped, 0 failed.

Note:

• The first Shared.Tests --no-restore run in the fresh worktree hit the repository-documented first-run project.assets.json gotcha.
• I restored the test projects and then repeated the full AGENTS-required sequence from the beginning: ./build.ps1, Shared.Tests, Tray.Tests.
• No private local paths, tokens, or endpoints are included in this proof summary.

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

[TheAngryPit]

@clawsweeper re-review

Correction and current-head proof for PR1.

I previously posted a re-review request for the stale PR1 head 0555bfa52c0de1d6b88055260d24995d2c2b7782. That was wrong because the real local PR1 base for the stacked PR2/PR3 branches also included the later validation-tooling commits. PR1 has now been pushed to the correct head:

a81719ba52d5f1b1cfce39f8fd64a446b202c230

What changed since 0555bfa:

• Hardened build.ps1 Windows/architecture/SDK detection for PowerShell hosts where $env:OS or $env:PROCESSOR_ARCHITECTURE is absent.
• Added scripts/Test-WindowsLocal.ps1, a local Windows validation lane that can run from a short subst path for deep checkouts.
• Kept GitVersion repository trust process-local during validation instead of mutating global Git config.
• Documented the local Windows validation lane in docs/TEST_COVERAGE.md.

Current-head validation on Windows Release Preview 25H2 build 26200.8728:

• ./build.ps1 -> passed; all builds succeeded.
• dotnet test ./tests/OpenClaw.Shared.Tests/OpenClaw.Shared.Tests.csproj --no-restore -> passed; 2262 passed, 29 skipped, 0 failed.
• dotnet test ./tests/OpenClaw.Tray.Tests/OpenClaw.Tray.Tests.csproj --no-restore -> passed; 1088 passed, 0 skipped, 0 failed.
• openclaw-autoreview --mode branch --base upstream/main --engine codex --model gpt-5.5 --thinking high -> clean; no accepted/actionable findings; overall patch is correct (0.82).

Safe claim:

• PR1 is current-head build/test/autoreview proven for the locale/test/local-validation baseline.
• PR1 still does not claim MXC runtime behavior; that remains in Validate MXC SDK 0.7.0 on Windows build 26200.8728 #786 and Add Gateway system.run MXC runtime E2E proof #787.

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

[TheAngryPit]

@clawsweeper re-review

PR1 was narrowed again in response to the previous review's merge-risk note.

What changed:

• Removed the local validation-helper/build-automation scope from PR1.
• PR1 now contains only the locale/test/formatter baseline:
  • invariant compact token formatting
  • OS-resolved known-folder test assertions
  • AppVersionInfo shared-state test serialization
  • explicit shared test-project marking
• The dropped validation-helper work is preserved only locally for now and is not part of this PR.

Current PR1 head:

0555bfa52c0de1d6b88055260d24995d2c2b7782

Current-head validation on Windows Release Preview 25H2 build 26200.8728:

• ./build.ps1 -> passed; all builds succeeded.
• dotnet test ./tests/OpenClaw.Shared.Tests/OpenClaw.Shared.Tests.csproj --no-restore -> passed; 2262 passed, 29 skipped, 0 failed.
• dotnet test ./tests/OpenClaw.Tray.Tests/OpenClaw.Tray.Tests.csproj --no-restore -> passed; 1088 passed, 0 skipped, 0 failed.
• openclaw-autoreview --mode branch --base upstream/main --engine codex --model gpt-5.5 --thinking high -> clean; no accepted/actionable findings; overall patch is correct (0.86).

Safe claim:

• PR1 is current-head build/test/autoreview proven for the narrow locale/test baseline.
• PR1 no longer includes the validation-helper automation surface that created the previous merge-risk note.
• PR1 still does not claim MXC runtime behavior; that remains in Validate MXC SDK 0.7.0 on Windows build 26200.8728 #786 and Add Gateway system.run MXC runtime E2E proof #787 after they are rebased on this narrowed base.