[arg274]

The EZ CD parser was written against one 10.3.0 log and patched for 6.0.0. It now follows a static analysis of how 2.8, 6.0, 7.1, 8.2, 11.5 and 12.4 write their rip logs, and reads every layout they produce in both the Normal/Range and Single File/Disc Image profiles.

What changes

• Tracks come from the Details blocks instead of the summary table. Cancelled and failed tracks are aborted, including a 12.4 track whose encoder fails to close and which logs no closing line. Both CRC label schemes are read (CRC / CRC verify up to 6.0, CRC (test) / CRC (copy) from 7.1).
• Per-sector errors are read in both formats (one line per flag up to 7.1, one combined line from 8.2), with positions merged into spans of adjacent sectors:
  • read → read
  • data (C2) → damaged sectors
  • synchronization → jitter
  • CRC (copy and test reads disagree) → inconsistency in error sectors
• TOC is built from sectors, covering all three table layouts and the reduced image log with no MSF columns. Pre-emphasis used to be read from the HDCD column on 12.4, and De-emphasized never matched.
• Settings newly mapped: AccurateCDDA → accurate stream, "Drive read cache" (2.8–7.1, inverted) → defeat audio cache, digital silence deletion, and 12.4's Loudness/Peak/Gain normalization, which used to leave normalization unknown. An album line without an artist keeps its title.
• Judgement calls, open to overruling:
  • Read mode is Burst for every mode, since none re-reads a sector.
  • "Detect pre-gaps: Enabled" means Append (except HTOA) and Disabled means Append (gaps undetected). Versions without the line give Unknown instead of Inapplicable, which triggered an incorrect gap handling deduction.
  • Null samples used is True. The analysis confirms this for 12.4 only.
  • The encoder is now the output extension (flac), as EAC and XLD report it, instead of the plugin text (FLAC 1.4.2).
  • The Range profile marks its tracks as ranges.
• A second commit renames new_xld's arguments and XLD's extract_errors locals from r_c, jg_c, m_s… to the field names, matching new_eac and new_ezcd.

Verification

• The real 10.3.0 log and two real 6.0.0 logs parse with matching CRCs. All three are clean rips.
• Error, cancel and failure handling was checked only against a 12.4 mock and synthetic 2.8, 7.1 (Range), 8.2 (image), 11.5 and 12.4 (reduced image; loudness plus silent encoder failure) logs built from the analysis. These parse the same with LF line ends and trailing spaces stripped.
• All 608 XLD sample logs give the same output as main, ignoring array order. Byte-level differences come from the HashSet-ordered encoder list, which varies between runs of the same binary.
• Clippy is clean on both commits.

🤖 Generated with Claude Code