Files
erofs-freebsd-out-tree/docs/TEST_REPORT.md
T
2026-08-18 09:20:44 +02:00

98 lines
3.7 KiB
Markdown

# Manual Test Evidence
## Final Status
repo22 has 161 executable manual test cases, TC001-TC161, plus the non-executable
TC000 template. The canonical dated reports produce:
| Status | Count |
| --- | ---: |
| PASS | 160 |
| PARTIAL | 1 |
| FAIL / KERNEL-FAIL | 0 |
| ENVIRONMENT-UNAVAILABLE | 0 |
| SHELVED test case | 0 |
TC146 is the single PARTIAL case. Its HEAD2 and interlaced paths pass, while
the explicit mapped-payload positive fixture remains unavailable. That
shelved subitem is not counted as a second test case.
## Evidence Policy
Authoritative evidence consists of the numbered Markdown procedures,
deterministic field-asserting fixture helpers, native syscall/kernel probes,
and dated `tests/results/manual/*/manual-test-report.md` reports. Reports record
the exact source, module or fixture hashes, FreeBSD kernel behavior, errno, and
cleanup state.
Host-only fixture generation, static ABI review, and userspace erofs-utils
output are supporting evidence. They do not replace a required FreeBSD kernel
result. The suite is manual-only; no CI pipeline or automated kernel-test
harness is implemented or claimed.
## Canonical Runs
The final status uses these non-overlapping report groups for TC001-TC156:
| Group | Exact scope | Result |
| --- | --- | --- |
| G1 | 32 IDs | 32 PASS |
| G2 | 13 IDs | 13 PASS |
| G3 | 31 IDs | 31 PASS after the dated TC060 fixed-source rerun |
| G4 | 27 IDs | 27 PASS |
| G5 | 24 IDs | 23 PASS, TC146 PARTIAL |
| G6 | 11 IDs | 11 PASS |
| G7 | 11 IDs | 11 PASS |
| G8 | 7 IDs | 7 PASS |
The G1-G8 tables contain exactly 156 rows and 156 unique IDs, with no missing
or duplicate TC from TC001-TC156. TC060's original G3 KERNEL-FAIL is historical
and is superseded by
`tests/results/manual/2026-08-09T1120Z-tc060-fix/manual-test-report.md`.
TC153 passed in the final G3 run and its issue is resolved.
TC157-TC161 are recorded in
`tests/results/manual/2026-08-09T1804Z-final-review-independent/manual-test-report.md`.
All five pass on exact source baseline
`fcc85b93d5f8fd9671686bd68bf3b086c8bd25cf`.
## TC111 Coverage Audit
A bounded audit of `tests/TC[0-9][0-9][0-9]-*.md` found 162 rows and 162 unique
IDs, TC000-TC161, with no missing or duplicate ID. Excluding TC000 leaves 161
executable specifications.
The audit also cross-checked the eight canonical group tables and the five
final-review rows. Final arithmetic is `155 PASS + 1 PARTIAL + 5 PASS`, or
160 PASS and one PARTIAL.
## Final Build Qualification
The final independent run rebuilt both configurations after the TC161 failing
`nm` shim and used `/usr/bin/nm` for the post-shim build:
| Configuration | KLD SHA256 |
| --- | --- |
| `WITH_ZSTDIO=0` | `15fda9d334132cd81769ce4dff4f8411a6e2b4cf7531c352530d0de85f42a2d2` |
| `WITH_ZSTDIO=1` | `23782dc0ce7da188807d35020bf2d8c6044b796c5c9a8746b398ba784cd6ad4e` |
Both KLDs loaded, mounted a qualified image read-only, unmounted, detached the
md provider, and unloaded. Final guest mount, md, EROFS KLD, and DTrace KLD
counts were zero.
G1-G8 evidence was collected across multiple source commits. It is not claimed
that all historical tests ran on the final KLD. The final code baseline was
independently exercised by TC157-TC161 and the affected dual-build,
module-load, and mount smoke.
## Issue Status
- TC010: RESOLVED by a real 16 TiB-plus sparse-provider statfs run.
- TC060: RESOLVED by the FreeBSD pathconf fix and exact-source rerun.
- TC153: RESOLVED by the multi-TiB Layout 0 directory lookup run.
- Explicit mapped payload: SHELVED as a detailed fixture/tooling limitation;
TC146 remains PARTIAL.
See `issues/README.md` for the current index and the individual files for
triggers, analysis, attempts, results, feature impact, and acceptance criteria.