v2.0.0

Added

• Clear error messages when a TBL file is malformed (truncated, offsets pointing past the end of the file, offsets in the wrong order), in place of crashes or silently incorrect output.
• Clear error messages when a text file being encoded contains an invalid <…> control sequence (missing >, non-numeric content, or values outside 0–255).
• An error is now reported if an encoded TBL would exceed the 64 KiB size limit, instead of silently producing a corrupt file.
• The analyse command now also reports offsets that point past the end of the file or that go backwards.

Changed

• The command-line interface has been reorganised around subcommands. Use irontbl tbl-to-txt <file>, irontbl txt-to-tbl <file> -o <out>, irontbl analyse <file>, or irontbl completions <shell>. The input file is now given as a positional argument rather than via -i/--input.
• Warnings and status messages are now written to standard error, so they no longer leak into piped output (for example, when redirecting generated shell completions to a file).
• Warnings about entries missing a trailing <0> now use the same 0-based entry numbering as the --line-number option.
• Malformed <…> control sequences in text input are now reported as errors instead of being silently dropped from the output.
• Characters outside the single-byte range (e.g. €, →) in a text file being encoded are now reported as errors instead of being silently truncated to a different byte. Latin-1 characters in the range 0x80–0xFF continue to round-trip correctly.

Removed

• The --mode, --input and --generate-shell-completions flags (replaced by the new subcommands).