Multiple improvements: seekable tests, zstdless CLI, streaming dict example, seqBench hardening, DiB early termination

[BhavyaBibra]

Summary

A batch of improvements across tests, examples, documentation, and contrib tools. Each change is independent and addresses existing TODOs or gaps in the codebase.

Changes

1. Seekable format: add unit tests & fix FIXME (contrib/seekable_format/tests/seekable_tests.c)

• Added 4 new test cases:
  • Zero-byte round trip — compress/decompress empty input
  • Checksum-enabled round trip — exercises checksumFlag=1 path
  • Out-of-bounds offset — decompress at offset past data returns error/zero
  • Invalid frame index — getFrameCompressedSize() etc. return errors for bad indices
• Replaced /* Github issue #FIXME */ with descriptive regression test comment (variation of ZSTD_seekable_decompress() can hang #2335)
• Removed /* TODO: Add more tests */

2. Improve zstdless script (programs/zstdless)

• Added --help / -h with usage message
• Added --version / -V to print zstd version
• Added ZSTDLESS_FLAGS environment variable for passing custom flags to zstd
• Removed the TODO comment

3. Add streaming dictionary compression example (examples/)

• New streaming_dictionary_compression.c demonstrating ZSTD_CCtx_loadDictionary() + ZSTD_compressStream2()
• Fills the gap between existing dictionary_compression and streaming_compression examples
• Updated Makefile and README.md

4. Document memset engineering decisions (lib/compress/zstd_compress.c)

• Replaced two /* TODO: avoid memset? */ comments (LDM hash table and bucket offsets) with documented analysis:
  • LDM is only enabled with --long (rare default path)
  • Generation counter approach would add branch overhead to every LDM lookup
  • memset cost is dominated by table allocation

5. Early termination in DiB_fileStats() (programs/dibio.c)

• Added early break when totalSizeToLoad >= MAX_SAMPLES_SIZE (2GB)
• Avoids unnecessary stat() syscalls when training dictionaries on large file sets
• Resolves /* TODO: there is opportunity to stop DiB_fileStats() early */

6. Harden contrib/seqBench (contrib/seqBench/seqBench.c)

• Added fopen() error check (previously segfaulted on missing files)
• Added all malloc() error checks
• Added fread() return value validation
• Added proper resource cleanup via goto cleanup
• Improved usage message with description
• Added compression ratio output and decompression size validation

Testing

• make -C lib libzstd.a ✅
• make -C contrib/seekable_format/tests test — all 9 tests pass ✅
• make -C examples all — all 10 examples build ✅
• make -C programs zstd — full binary builds with dibio.c changes ✅
• seqBench compiles successfully ✅
• zstdless --help prints usage ✅