# Test Case: MicroLZMA Consumption and Corruption

**Test ID**: TC144-microlzma-consumption-and-corruption

**G5 fixture contract**: Use [G5-MANUAL-SETUP.md](G5-MANUAL-SETUP.md).
Its generated paths, source comparisons, structured corruption offsets, and
cleanup rules supersede placeholder examples in this file.
**Category**: Compression
**Priority**: Critical
**Regression**: MicroLZMA full/partial completion semantics

## Objective
Verify MicroLZMA input-consumption rules for complete streams, early-success
rules for partial references, and deterministic rejection of damaged streams.

## Preconditions
- FreeBSD 15 guest and repo22 module.
- A full-index LZMA image with:
  - `/a.dat`: 1048576-byte complete MicroLZMA stream;
  - `/b.dat`: 700000-byte partial reference to `/a.dat`'s pcluster;
  - the `/b.dat` HEAD record marked `Z_EROFS_LI_PARTIAL_REF`.
  - the reused pblk and `D0_CBLKCNT` both equal `/a.dat`'s values.
- A corrupted copy with a byte changed inside the proven compressed byte range.
- Source files and their SHA256 values available on the guest.

## Test Steps
1. Prove the pcluster and partial-reference record with `dump.erofs` and a
   byte-level parser.
2. Mount the valid image and hash `/a.dat` completely.
3. Hash `/b.dat` completely; this requests only a prefix of the reused stream.
4. Compare reads at offsets 65500 and 524287 against source bytes.
5. Mount the corrupted image and read `/a.dat` with an errno-reporting helper.
6. Confirm the implementation's complete-stream condition includes both
   `XZ_STREAM_END` and `buffer.in_pos == srclen`.
7. Compare `vmstat -m` active EROFS allocations and pre/post `dmesg`.

## Expected Results
- `/a.dat` succeeds only after exact output, `XZ_STREAM_END`, and complete
  compressed-input consumption.
- `/b.dat` accepts `XZ_OK` after the requested partial output is filled.
- Both complete SHA256 values and random ranges match.
- The corrupted full stream returns errno 5 (`EIO`).
- `control.bin` in the same corrupted image still matches its source.
- Decoder state is freed on success and error; active EROFS allocations return
  to zero after unmount/unload.

## Cleanup
```sh
umount /mnt/repo22 2>/dev/null || true
mdconfig -d -u "${unit#md}" 2>/dev/null || true
kldunload -i "${module_id}"
```