Main now has its own proxy class handling for DWG saves (`proxy_classes`, from bf137e2c,
31935157 and 946af874) and the application object conversion of #2917. This branch is
brought onto that:
- Cargo.lock, crates/dwg/Cargo.toml and xtask/src/layers.rs are main's. `dwg` no longer
depends on `dxf`: the source classes are read from the bridge's ASCII DXF by a bounded
scan in `proxy_transport::source_classes`, as main's `dxf_class_names` reads their names.
- Saving as DWG 2013 or later restores the source classes in their places (duplicate
names and metadata kept) when every CLASS record can be written back. Otherwise (before
2013, a binary DXF, a malformed or unused bad CLASS) main's `proxy_classes` runs: proxies
are matched by name or left out and counted as unwritten. Nothing returns an error any
more, so a drawing with proxies saves in its own pre-2013 version, without them, as the
save report says. A proxy whose number names no source class is left out and counted.
- The loss report is main's `not_saved_as` (`Foreign::proxy_class`);
`preserves_proxy_entity` and this branch's report test are gone.
- On DWG read, `prepare` runs after `application_objects`, so the proxy objects that makes
are numbered as the DXF numbers them. It uses main's `proxy_dwg_version`. A registered
entity with decoded properties and opaque data now stays as read, as on main, and a read
note names it, instead of being dropped.
Tests: `cargo test -p cadcraft-dwg` 37 passed, `-p cadcraft-io` all passed (25 ignored),
`-p cadcraft-doc`, `-p cadcraft-engine --test save_reports_not_saved` and `--test it`
passed; clippy `-p cadcraft-dwg -p cadcraft-io -p cadcraft-doc -p cadcraft-engine
--all-targets -D warnings`, fmt, `cargo xtask layers` and the wasm32 check of cadcraft-io
pass. `registered_proxy` now checks that a pre-2013 DWG save leaves the proxy out and
keeps the rest; main's proxy tests in crates/io/src/tests.rs pass unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H43qQA41CJLauqXnPgBQ7r
hard_owners took the first object naming a null-owner object in a 360 as its owner, kept or not.
When an object that is left out named it first, the reader dropped the object, though a kept one
held it too, and the save wrote the kept holder's 360 as 0.
hard_owners now lists every object naming it, in file order, and owner_of takes the first of
them that is kept, else the first. The read (retained_objects) and the write liveness pass in
finish run until what they keep settles; a file where no object has two hard owners takes one
pass, as before. The write pass no longer keeps a converted copy of every object's groups.
The test adds an object nothing keeps that names the held record before its holder does: the
held record stays with the holder and the stray object is dropped, DXF and DWG, twice. It fails
before this change ("DXF 1: no record held").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H43qQA41CJLauqXnPgBQ7r
About > Contributors still showed an old snapshot, because nothing
regenerated contributors/contributors.json before a release. Opt in to
the shared refresh from craft-common-actions: a read-only credits job
regenerates the snapshot for the release commit, and the shared release
workflow installs it before every source build. The snapshot is not
committed back.
An object whose owner (group 330) is null but which another kept object
holds by a hard owner reference (group 360), as one application's proxy
object held by another, broke the owner chain the reader and writer follow
to decide what to keep: it was dropped, and the holder kept a reference to
nothing. Such an object's owner is now the object that hard-owns it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H43qQA41CJLauqXnPgBQ7r
An object of another application's class that acadrust has no model for is
read only as its DWG record (ObjectType::Unknown), which its DXF writer
leaves out. The DWG to DXF bridge lost every such object with no note, and
the references to it went too (an extension dictionary entry naming the
object, say).
Each one is now carried as a proxy object of its class (ACAD_PROXY_OBJECT):
the record is split into its common object data (owner, reactors, extension
dictionary) and the streams after it (data bits, R2007+ strings, object
references), as the Open Design Specification for .dwg describes them. The
references are read to the end of the handle stream (to its exact length
from R2010 on), null ones too: a null reference is an empty field, and the
references after it belong to the next fields. Objects of the drawing
database's own classes, and records that don't split, stay as they were and
are named in a note.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H43qQA41CJLauqXnPgBQ7r
Without a path the screenshot was written into a private temp directory, and std::env::temp_dir panics on wasm32 (no file system), which stopped the whole web app: the page's call never got a reply and nothing after it ran. On wasm the PNG now comes back inline (pngBase64, width, height, bytes), as ui.render's does. Native builds are unchanged.
The web shell queues the page's calls as control requests (the channel the desktop control server uses) and answers them on the next frame. PendingReply gains try_reply, so a transport that cannot block can poll it, and cancels a request the app has not started when it is dropped, as a timed-out wait does: the page was told the call failed, so the app skips it. The API is installed in every mode; opening, saving, the file name and the unsaved flag stay with the host messages.
- #2587: a frame turned by anything but a quarter turn keeps its size about its turned
centre; the box of its two mapped corners collapsed it (a 10 x 10 square turned 45° became
14.14 x 0, and the OLE header with it) for EXPLODE, REFEDIT, paste and array placement
(austin62jake, CodeRabbit). Moves, scales, mirrors and quarter turns keep the exact
corner mapping. `recase` reads only the rewritten digits and a few after them, not the
whole 310 line (austin62jake). The PR's rotated-frame test expects the kept size.
- #2585: an `output` that contradicts the path's extension is refused, as a device that
does is; it lost silently to the extension.
- #2608: the window's state is dropped when another dialog took its place (it missed a UI
pass), so reopening reads the drawing instead of an old unsent edit (CodeRabbit); TEXT and
ATTDEF get a one-line field (Enter commits), so a line break can't turn into a space on
save; the module comment says the shortcut menu opens it.
- #2471: an unnamed CLASS record keeps its place in the class list a DWG maps proxies by,
so the classes after it keep theirs; a proxy of the unnamed class is left out instead of
taking the next class's name (the ODA refuses such a DWG).
- #1626: the first *ACTIVE VPORT record is the caller's to say. A first record with a
usable view but an unusable tile kept no tile, so the next *ACTIVE record was taken for
the first and its view replaced the saved one (also austin62jake on #1626).
- #2584: a tilted LWPOLYLINE/HATCH run is flattened to at most a million vertices, then
its arcs stay chords (a hostile file of arcs asked for gigabytes; austin62jake). The
file-format-parity.md hunk is dropped: maintainers update the parity docs.
- #2471: another program's proxy objects are in the DWG loss report as its proxy entities
are (left out before 2013, or when their class can't be named); they went silently.
- #1626: a drawing saved at a view keeps it in the desktop app and when it has xrefs: the
app's first frame and the xref pass both fitted the extents over it. OPEN's result says
`saved_view`; one `view::saved_view` replaces the copy of `usable_view` in files.rs.
A misplaced doc comment goes back to its test.
- #2028: a plot rotation (73) outside 0-3 is written back as 0, not as read.
- #1026: the binary code-page retry drops the first pass's tags before it runs (the peak
doubled near the tag limit); a float spelling of an integer group clamps to what the
width holds unsigned too, so 65535.0 reads as 65535 does, not 32767.
- #2658: the reference names Undo by the last command ("Undo Group of commands"); Undo
matched nothing and went to the end of Edit. Ordering and its test share one rule.
- #2612: its test's Insert gets `clip: None` (XCLIP landed after it was written).
FredApps