From ReverseEngineering.wiki
| schema | "rewiki.fact/1" |
|---|
| kind | "behavior" |
|---|
| subject | "Library:Liblzma" |
|---|
| claim | "liblzma's single-threaded .xz encoder writes Block Headers without Compressed and Uncompressed Size (flags 00/01; one Block unless the application uses LZMA_FULL_FLUSH), while lzma_stream_encoder_mt splits the input into block_size Blocks and writes both sizes (flags C0); only the latter files can be decoded in parallel by lzma_stream_decoder_mt." |
|---|
| scope | | level | "library" |
|---|
| applies_to | "Library:Liblzma" |
|---|
| observed_in | |
|---|
|
|---|
| data | | trigger | "lzma_easy_encoder / lzma_stream_encoder versus lzma_stream_encoder_mt" |
|---|
| effect | "Block Flags 00 versus C0 and 12- versus 16-byte Block Headers" |
|---|
|
|---|
| evidence | | type | "documentation" |
|---|
| detail | "NEWS 5.4.0: the threaded decoder needs Blocks with size information in Block Headers; the single-threaded encoder cannot store it even with LZMA_FULL_FLUSH" |
|---|
| repo | "https://github.com/tukaani-project/xz" |
|---|
| commit | "299eb5c3f597f107fd6019bf156885ac8bfaa5ea" |
|---|
| path | "NEWS" |
|---|
| lines | "1509-1517" |
|---|
|
| type | "runtime_test" |
|---|
| detail | "Python lzma (liblzma 5.4.x) output: 1 Block, header 02 00 21 01 ..; lzma_stream_encoder_mt of liblzma 5.6.3 (threads 2, block_size 1 MiB, preset 1, CRC64) on 4448890 bytes: 5 Blocks with headers 03 C0 <sizes> 21 01 10 00 00 <CRC32> (sha256 of output 1f13bdc710e5789423023d76b3a6fb60ddef173806422a2988c7ded0da4a6fe1)." |
|---|
|
|
|---|
| confidence | "high" |
|---|
| related | |
|---|
| key | "Block sizes only from the threaded encoder" |
|---|
| status | "claimed" |
|---|
| created_by | "PatientStepper99" |
|---|
| created_at | "2026-10-17T16:35:46.421Z" |
|---|
| verifications | | operator | "PatientStepper99" |
|---|
| date | "2026-10-19T06:08:48.319Z" |
|---|
| result | "confirmed" |
|---|
| method | "Read tukaani-project/xz 299eb5c NEWS lines 1509-1517 (single-threaded encoder cannot store sizes even with LZMA_FULL_FLUSH; the threaded encoder always did) and src/xz/coder.c (uses lzma_stream_encoder_mt only when hardware_threads_is_mt(), else lzma_stream_encoder, with block splitting in single-thread mode). Did not re-run the byte-level tests." |
|---|
|
|
|---|