[codeforester]

Summary

• Preflight manifest-command trust for every selected workspace test project before starting any delegated test.
• If trust preflight is blocked, start zero project test commands and report blocked versus not-run projects explicitly.
• Add text progress for preflight, start, completion, elapsed time, and dry-run planning.
• Publish workspace-test JSON schema version 2 with explicit preflight, execution, blocked, and not-run state.

Issue

Closes #2612

Validation

• env -u BASE_HOME BASE_TEST_PYTHON=/Users/rameshhp/.base.d/base/.venv/bin/python BASE_BASH_LIBS_DIR=/Users/rameshhp/work/base-bash-libs/lib/bash ./bin/base-test — 1,826 Python tests and 1,052 Bats/integration tests passed.
• Focused workspace/trust/schema tests: 41 passed, 18 subtests passed.
• Live blocked-workspace reproduction reported 0/7 project test commands started and marked trusted projects NOT RUN.
• Stable compatibility check passed.

Notes

The PR is intentionally not merged by this train; merge requires explicit approval.

[codeforester]

Automated review findings

1. --dry-run can report false success for a selection guaranteed to fail (cli/python/base_projects/workspace_test.py ~line 452): dry_run_workspace_test_results reports every action=="skip" target as status="skipped" unconditionally, ignoring target.fatal. The real execution loop a few lines away computes status = "failed" if target.fatal else "skipped" for the identical case. Confirmed directly against both code paths. Concretely: a workspace manifest selects a required repo that's missing/outside the workspace (inspect_workspace_repo returns a fatal skip state). basectl workspace test --dry-run --format json reports that project as "skipped" and exits 0, because the dry-run exit code (FAILURE if (preflight_blocked or preflight_failed) else SUCCESS) never looks at per-target fatal status. Dropping --dry-run for the identical selection makes counts.failed nonzero and the command exit 1 - the dry run promises success for a selection that's guaranteed to fail, defeating its purpose of previewing the real outcome.

2. Redundant manifest re-parsing (~line 379): workspace_test_preflight re-reads and re-parses every selected project's base_manifest.yaml, which was already parsed moments earlier by inspect_workspace_repo during target construction - and the manifest is parsed a third time inside each delegated basectl test subprocess. For a workspace with N selected test projects, this is 2-3x the manifest I/O/YAML parsing needed, scaling linearly with workspace size for no behavioral benefit. Passing the already-parsed manifest through WorkspaceTestTarget (or caching it) would avoid the duplicate work.

3. Reuse gap: workspace_test_preflight reimplements the trust-identity-lookup logic that already exists in base_projects/test_preflight.py::project_test_preflight, instead of extracting a shared, batch-capable helper. The trust-identity/trust-status computation now has two independent call sites with different detail-message formatting (this one's "manifest command trust is blocked (...)" doesn't match the wording in manifest_trust_guidance.print_blocked_command_text) - a future change to trust semantics has to be applied in both places and risks drifting further.

Posted via Claude Code

[codeforester]

Addressed the automated review findings in ed66295.

• Dry-run now preserves fatal required-project selection failures as status=failed and exits non-zero; added a regression test for a missing required repository.
• WorkspaceTestTarget now carries the parsed manifest from workspace inspection, so workspace preflight no longer reparses base_manifest.yaml.
• Extracted project_test_trust_status() and reused it from both project test execution and workspace preflight, with one trust-store instance for the workspace batch.

Validation: full local gate passed (1,827 Python tests and 1,052 BATS tests), stable compatibility passed, targeted pylint passed, and git diff --check passed. PR remains open and unmerged.