CLA-11-03 deep review — September 29, 2026

The study guide maps all 23 numbered objectives, grouped 6/7/6/4. All 30 original answers were expanded. The active July 24, 2025 syllabus retains weights 29/38/25/8 and item counts 12/14/10/4. Manual comparison followed an automated timeout; existing snapshots were retained. This is same-context AI review with human review pending.

Evidence and delivery

The complete exam page, catalog, policies and Pearson scheduling page were read through public fallback. CLA's 15-day failed-retake rule is separate from entry-level TestNow. Newer scheduling copy explicitly includes CLA in its OnVUE overview; conflicting anytime-before/24-hour rescheduling language remains qualified. No actual appointment or account was inspected.

31 sources are registered; direct health outcomes are {'error': 9, 'ok': 20, 'blocked': 2}. Individual boundaries distinguish complete landings, selected primary sections, web fallback, shells and inaccessible content. Selected historical WG14 draft clauses and SEI/GCC references support the repairs; no full language-standard audit or local PDF capture succeeded. C17 is an explicit practice choice. Both public Edube landings list 42 hours; modules, historical CLA alignment and inconsistent course wording are qualified. Microsoft’s C89-organized index and community navigation are not substituted for portable language contracts. Paid interiors and rendered Cisco lessons were unavailable.

Repairs and execution

The old allocation fragment referenced an undeclared variable. The positive-delta function could overflow signed int, and the formatted-input loop obscured matching failure and unchecked close behavior. The guide now supplies five complete original C files and an optional CMake build: a guarded API, bounded record loader, command-line application, 181-check loader driver and 53-check language workbook. The loader distinguishes text, I/O, allocation and count failures, rejects NUL/overlong lines, accepts final lines without newline, and preserves existing caller data on failure. Allocation dimensions are checked before multiplication; successful realloc invalidates old aliases. Fictional limits and parser grammar are explicit.

Eleven final remote build/run receipts passed. GCC 15.2 and Clang 20.1 each passed 181 loader checks and 53 language checks. Selected GCC optimized/configuration builds and a Clang AddressSanitizer/UndefinedBehaviorSanitizer build passed. Application success/malformed cases create fixtures within the execution sandbox and call the unchanged application body through a small driver; ordinary entry-point usage/missing-file cases were also run. Supplied build text files were not present in the execution sandbox, so the fixture method was corrected before final verification. No local compiler was installed.

Six negative checks observed expected preprocessing/compile/link rejection, including the former allocation error and duplicate definitions across translation units. The duplicate-definition response had top-level code zero but a failing build step; the harness was corrected to inspect step codes and actual execution separately. A preprocessing-only request produced 457 lines with the extended configuration. No undefined runtime example was executed. All six public files were compared to final request content and hashes; full private receipts plus concise public evidence preserve the execution boundaries.

The byte reader enforces UCHAR_MAX <= INT_MAX at compile time. Test fixtures use exclusive creation and cleanup. Injected allocation failure is not real exhaustion; low-level stream read/close/output-flush failures and an interactive debugger were not exercised. Sanitizer success covers tested paths only. Eight broader learner activities remain proposed. Source examples containing malformed or incomplete ownership/boundary logic were not copied.

The operation record preserves objective mappings, reading boundaries, public code and remote execution evidence. Full unit, repository, catalog and strict generated-site gates are recorded after completion. The outcome is reviewed-with-blockers for source/appointment limits, unexecuted broader work and pending human review. No installation, persistent service, auth/permission change, account, purchase, exam attempt or external message occurred.