fix: stop the portability audit from failing renders over their own tests

The final audit walked the whole build folder and treated any codeplain-tty
reference as a violation, on the stated assumption that internal tests live
outside it. They do not: a module's build folder carries a conformance_tests/
subtree, and those tests drive the helper because that is what it is for.

So CreateDist raised PlatformBoundaryViolation *after* the render succeeded.
The build was never copied, run_state.render_generated_code_path stayed empty,
and the harness fell back to whatever tree it could find. Run against the
cli-password-manager artifact from codeplain-tty-capability-run2, the audit
flags four files, all of them conformance tests — which is why that render
reported 22 functionalities in 48m54s and delivered a build with no CLI entry
point, and why code_retrieval_example scored 0/10 with generated_code=- in the
wave2 re-baseline.

Two changes:

- Internal test trees (conformance_tests, acceptance_tests,
  dist_conformance_tests) are exempt from the audit. Delivered code beside them
  is still audited — a test asserted for that explicitly.

- The exit summary and the render trailer now report an error whenever there is
  one, not only when the render failed. A render that completes its
  functionalities and then raises on the way out printed a success banner,
  logged no reason at all, and exited 1 — which is exactly how this went
  unnoticed across a whole benchmark run.

Author: gorandodig

cli_output/render_summary.py

@@ -35,7 +35,11 @@ def print_exit_summary(
     msg += format_usage_summary(run_state.rendered_functionalities, run_state.render_time_accumulated) + "\n"
     console.print(msg)
 
-    if not run_state.render_succeeded and error_message:
+    # Reported whenever there is one. A render can finish its functionalities and still
+    # raise on the way out — publishing the build, for instance — and that combination
+    # used to print the success banner and swallow the reason entirely, leaving a caller
+    # with a tick mark and a non-zero exit code.
+    if error_message:
         console.error(error_message)
     console.quiet = True
 
@@ -70,7 +74,7 @@ def log_render_trailer(
         f"generated_code={run_state.render_generated_code_path or '-'} "
         f"spec={spec_filename}"
     )
-    if outcome == "failed" and error_message:
+    if error_message:
         logger.error(f"{RENDER_TRAILER_PREFIX} error={error_message}")
 
     # The process may exit immediately after this; an unflushed trailer would defeat the

render_machine/platform_test_audit.py

@@ -7,9 +7,11 @@
 application must run in a clean environment where none of Codeplain's test tooling
 exists.
 
-Internal conformance and acceptance tests live outside the build folder (in the
-module's tests tree), so nothing here needs an allowlist: any hit inside the build
-folder is a violation.
+Internal conformance and acceptance tests are exempt. They are what the helper exists
+for, and they do turn up inside the audited tree — a module's build folder can carry a
+`conformance_tests/` subtree, and benchmark renders showed the audit failing them.
+Auditing those is not a stricter boundary, it is a false one: it aborts a successful
+render over test code that is never delivered.
 """
 
 import os
@@ -22,6 +24,11 @@
 # Directories that carry no delivered source and may be large.
 SKIPPED_DIRECTORIES = {".git", ".venv", "node_modules", "__pycache__", ".tmp", "dist", "build", "target"}
 
+# Internal test trees, which are allowed to drive the helper and are never delivered.
+# Named separately from the above because skipping them is a boundary decision, not a
+# performance one.
+INTERNAL_TEST_DIRECTORIES = {"conformance_tests", "acceptance_tests", "dist_conformance_tests"}
+
 MAX_AUDITED_FILE_BYTES = 4 * 1024 * 1024  # a delivered source file larger than this is not source
 
 
@@ -33,7 +40,9 @@ def find_platform_references(build_folder: str) -> List[str]:
     """Build-folder-relative paths of files referencing the platform test helper."""
     violations = []
     for root, directories, file_names in os.walk(build_folder):
-        directories[:] = [name for name in directories if name not in SKIPPED_DIRECTORIES]
+        directories[:] = [
+            name for name in directories if name not in SKIPPED_DIRECTORIES and name not in INTERNAL_TEST_DIRECTORIES
+        ]
         for file_name in file_names:
             path = os.path.join(root, file_name)
             relative = os.path.relpath(path, build_folder)

tests/test_platform_test_audit.py

@@ -43,3 +43,28 @@ def test_vendor_directories_are_not_audited(tmp_path):
     (tmp_path / "app.js").write_text("console.log('clean');\n")
 
     audit_build_folder(str(tmp_path))  # does not raise
+
+
+def test_internal_test_trees_inside_the_build_folder_are_exempt(tmp_path):
+    """A module's build folder can carry its conformance tests, and those are what the
+    helper exists for. Auditing them aborted successful benchmark renders at CreateDist:
+    the build was never published, `generated_code` stayed empty, and the delivered
+    artifact was whatever the harness could salvage."""
+    (tmp_path / "conformance_tests" / "init").mkdir(parents=True)
+    (tmp_path / "conformance_tests" / "init" / "test_init.py").write_text(
+        "subprocess.run(['codeplain-tty', 'wait-for', 'Password:'])\n"
+    )
+    (tmp_path / "conformance_tests" / "conformance_tests.json").write_text('{"codeplain-tty": true}\n')
+    (tmp_path / "vault.py").write_text("print('hello')\n")
+
+    assert find_platform_references(str(tmp_path)) == []
+    audit_build_folder(str(tmp_path))  # does not raise
+
+
+def test_delivered_code_is_still_audited_alongside_them(tmp_path):
+    """Exempting the test tree must not exempt the application beside it."""
+    (tmp_path / "conformance_tests").mkdir()
+    (tmp_path / "conformance_tests" / "test_init.py").write_text("codeplain-tty wait-for\n")
+    (tmp_path / "vault.py").write_text("os.environ['CODEPLAIN_TTY_ENDPOINT']\n")
+
+    assert find_platform_references(str(tmp_path)) == ["vault.py"]

tests/test_render_trailer.py

@@ -99,3 +99,14 @@ def test_the_trailer_is_flushed_so_it_survives_an_abrupt_exit():
         logger.removeHandler(handler)
 
     assert handler.flush.called
+
+
+def test_a_completed_render_that_still_raised_reports_the_reason(trailer_lines):
+    """A render can finish its functionalities and raise on the way out — publishing the
+    build, for instance. That combination printed a success banner, logged no reason, and
+    exited non-zero, which is how a boundary-audit failure went unnoticed across a whole
+    benchmark run."""
+    lines = trailer_lines(run_state(succeeded=True), error_message="The generated build references ...")
+
+    assert any("outcome=completed" in line for line in lines)
+    assert any("error=The generated build references ..." in line for line in lines)