CHDSharp v1.5.2

CHDSharp v1.5.2 - What's New

Released: 2026-10-21 - Full changelog - NuGet

CHDSharp v1.5.2 rolls up the v1.5.1 cue-layout parity fixes and adds a repo-wide code-quality pass plus fully automated wiki and
documentation-site deployment.

---

Download (CLI)

Platform	Bundle
Windows x64	CHDSharp_win-x64_v1.5.2.zip
Windows ARM64	CHDSharp_win-arm64_v1.5.2.zip
Linux x64	CHDSharp_linux-x64_v1.5.2.zip
Linux ARM64	CHDSharp_linux-arm64_v1.5.2.zip
macOS Intel	CHDSharp_osx-x64_v1.5.2.zip
macOS Apple Silicon	CHDSharp_osx-arm64_v1.5.2.zip

Every bundle contains a framework-dependent single-file CHDSharp / CHDSharp.exe
build (the .NET runtime is not embedded) plus LICENSE.txt, README.md, README.txt
and this document. The .NET 10 runtime
is required.

---

Fixed: REM LEAD-OUT at or before the session's last track

The parser compared the track's frame count with leadOut - index00 using two int values.
MAME compares a uint32_t frame count against a signed int32_t end frame
(cdrom.cpp:2748), so a negative end frame converts to a huge unsigned value and the trim is
skipped. Cues that declare REM LEAD-OUT 00:00:00 (or any address at or before the last track
of the session) therefore failed with:

Track 2: the session lead-out values leave a negative length (-11400 frames); the cue's
REM SESSION/LEAD-OUT values do not match the image.

while chdman createcd encoded the same cue, because the no-op lead-out leaves the cue's INDEX
positions in charge of the layout. The parser now performs the same unsigned comparison, so the
track layout and the CHD bytes match chdman.

Fixed: a pad larger than its track zeroed the whole track

The CD reader treated the last PadFrames frames of a track as zero-filled padding. In chdman
the pad start is an unsigned offset from the track start (chdman.cpp:471-472); when the pad is
larger than the track the offset wraps below the track start, the pad region is unreachable, and
every frame is read from the source file. The encoder now mirrors that rule, and the GD-ROM
data/pad/split ordering follows the same frame layout (splitStart = frames - splitFrames,
padStart = splitStart - padFrames).

Code quality (new in v1.5.2)

The whole solution is now clean under the ReSharper command-line inspector:
jb inspectcode CSharp_CHDSharp.sln --severity=WARNING reports zero warnings or errors
(308 before the cleanup). The policy lives in the root .editorconfig:

• Vendored codecs are exempt and untouched. VendoredZLib, VendoredLZMA, VendoredFlac
  and VendoredZSTD are third-party ports; byte-for-byte chdman parity depends on them staying
  verbatim, so their style/quality inspections are disabled instead of "fixed".
• Consumer-facing models are exempt from "never used" inspections - public library records,
  the JSON corpus manifest, and the WPF data models are read by callers, serializers and XAML
  bindings rather than by the declaring assembly.
• First-party findings were fixed: redundant null checks in the CLI encode paths, a write-only
  AVI stream field, and dead tester state (LogEntry/LogEntries and the unread
  PerFileResult.FilePath).

CI, wiki & docs (new in v1.5.2)

• Docs deploy from CI. After the three-OS build/test matrix passes on a push to main, the
  CI workflow regenerates the GitHub wiki from docs/ and requests a Pages rebuild. The release
  workflow does the same for tags, so the wiki and the documentation site stay in sync with the
  code.
• Dark theme by default. The documentation site uses a dark palette with a persisted
  light/dark toggle and color-scheme metadata.
• Badges everywhere. Every README (root, library, CLI, tests, generator, tester, benchmarks
  and docs/) starts with its CI/NuGet/platform badge row.
• Release routine documented. AGENTS.md now records the version bump, verification, tag,
  bundle and wiki/Pages steps used for this release.

Verification

• New chdman createcd byte-parity fixtures: zero lead-out, lead-out before the track, and the
  pad-larger-than-track layout.
• Parser regression tests for both lead-out shapes.
• CHDSharpEncoderTest 454/454, CHDSharpTest 4214/4214, battle harness --quick 905/905 checks.
• Full dotnet test across net8.0/net9.0/net10.0: 14,004 passed, 0 failed; full battle
  harness: 3,174/3,174 checks, byte-identical chdman output.

Upgrading

No public API changes. Replace the package; cues that encoded before produce the same bytes, and
the two layouts above now encode instead of throwing.