Francesco BVandClaude Opus 5.5 9107479ade io: open -> save -> reopen round-trip harness with loss report (FF5) ...

cadcraft_io::roundtrip compares a drawing with itself saved and reopened
(entities by handle, tables, blocks, layouts, unmodelled objects, xdata,
classes, header variables; numbers within a tolerance), and
roundtrip::corpus runs a folder of DXF/DWG files through DXF and DWG into
report.md and report.json; exposed as `cadcraft-cli roundtrip` and
`cargo xtask corpus`. The corpus itself stays outside the repo.

The DWG bridge now says what its writer leaves out, per DXF type
(cadcraft_dwg::dxf_to_dwg_noted, cadcraft_io::write_noted): the report
shows those records as lost on save and a strict corpus run fails on
them. OLE2FRAME, which the writer keeps with its data, is no longer
listed as not saved in a DWG.

CADCRAFT_CORPUS_STRICT=1 makes cadcraft-cli roundtrip (and so cargo xtask
corpus) fail on the same strict failures as the corpus test; xtask makes
a relative corpus folder and --out absolute before running the CLI.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

2026-10-27 17:45:34 +01:00

..

data

M0: CADCraft skeleton and vertical slice

2026-10-05 23:22:25 -04:00

src

io: open -> save -> reopen round-trip harness with loss report (FF5)

2026-10-27 17:45:34 +01:00

Cargo.toml

Use chrono instead of hand-rolled UTC conversion (#1789)

2026-10-22 14:09:16 -04:00