- Process ids are reused: report_for_pid now takes the sentinel's arm
time and ignores a report written before it, so an old crash of an
unrelated run with the same pid no longer costs a backend. The test
plants such a stale report.
- is_device_failure drops "buffer" and "surface": drawing code and the
modelling kernel panic with those words too. It keeps wgpu, out of
memory and device lost, and adds wgpu's "validation error".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#1519 stopped the sentinel blaming a backend for a run that had already
drawn for SENTINEL_PROOF_DELAY, on the reasoning that a backend which
kept the application alive that long had proved itself. That reasoning
does not hold for the failure this repository sees most on integrated
hardware.
Measured here on an Intel Iris Xe: with `--backend dx12` a large drawing
aborts the process 110 s in, long past the proof delay, so the sentinel
carried `survived=1` and the next launch went straight back to dx12 and
crashed again. The same binary on `--backend vulkan` opens the same
drawings with no out-of-memory report and no uncaptured errors at all,
so the backend really is the difference and moving off it really is the
cure. D3D12 is fine while the interface draws and only fails once the
whole scene uploads, which is why the clock is the wrong question.
The crash report from #1521 answers the right one. It records the
backend that was live and what the panic was, and it is named for the
process that wrote it, so it joins exactly to the `pid` the sentinel
already stores — no guessing from whichever report is newest, and no
new state to keep. A sentinel marked survived is now excused only when
that run left no report, or left one that has nothing to do with the
graphics device.
The test that matters is the pair, and both were run end to end:
killed from Task Manager after 40 s, no report -> dx12 kept
device crash after 110 s, report names wgpu -> dx12 skipped, Vulkan
`is_device_failure` is deliberately narrow: an indexing bug in drawing
code that happens to run on a frame must not cost the user a working
backend.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
#1521 inserted crash_log between web_diag's doc comment and its module, so the comment documented crash_log instead. #1519 wrote the survived marker with a literal line break inside the string.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A release build hides the console and is stripped, so a panic ended the
process with nothing on screen, nothing on stderr and nothing on disk.
The window simply disappeared. That is all a reporter has ever had to
work from, which is why #635 and #845 ask for crash logs and why a
crash like #830 can be reported for months with no detail beyond the
steps to reach it.
Every panic now writes one file to <config>/crash_logs naming the
version, the OS, the graphics backend that was live, the thread, the
panic message, the source location and a backtrace. The hook chains
onto whatever was installed before it, so a debug build still prints to
stderr as well, and it is installed after the child-process handoffs:
a GPU probe aborting is how an unusable backend gets detected, not a
crash worth filing.
The report is written by a plain function taking the message and
location, so the layout is covered by tests without installing a panic
hook -- a test that asserts inside a hook fires on every other test's
panic and aborts the whole binary, which is how the first version of
these tests failed.
Backtraces use `force_capture`, not `capture`: a report that says
"disabled backtrace" because the user never set RUST_BACKTRACE is
exactly the report that helps nobody.
The text is deliberately English and does not go through `t!`. It is
read by whoever receives the bug report, and a panic handler that calls
into the localisation machinery risks a second panic while handling the
first.
Confirmed on Windows 11 by reproducing a real crash: opening either of
two large drawings on an Intel Iris Xe dies inside wgpu, and the report
names it -- "Buffer with '(wgpu internal) StagingBelt staging buffer'
label is invalid", backend dx12, with the full frame list.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Capture the rendered canvas during the clean frame instead of requesting synchronous GPU buffer readback. Preserve embedded previews and allow saving when canvas readback is denied. Add a browser save and reopen regression check.
Fixes#1021
iced's clipboard rides window_clipboard, which has no wasm backend — a
focused text_input captured Ctrl+C/V on the web and did nothing.
Ctrl+V (captured, wasm only): read the async browser clipboard (the
keypress is a user gesture, so it's permitted) and replay the text as
synthetic KeyboardEvents on the winit canvas — the only route into a
focused text_input, since iced exposes no insert-text operation.
Control characters are dropped and the length is capped. The MText and
inline-TEXT editors keep their dedicated paste paths (no double feed),
and an unfocused Ctrl+V still runs PASTECLIP.
Ctrl+C (captured, wasm only): a widget operation pairs each text_input
with its focus state (iced fires both callbacks back-to-back on the
same widget) to collect the focused field's visible text, then writes
it with navigator.clipboard.writeText. Whole-field text, not the
selection — iced doesn't expose the selected range.
Native behaviour is unchanged.
Closes#346
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
wgpu/naga report pipeline and shader failures through the log facade,
then leave the canvas empty - with no logger installed on wasm the
message was lost, so a broken GPU path looked like "draws nothing" with
a clean console. Install a web logger that mirrors Error-level records
to the browser console and into a fixed DOM overlay with selectable
text, a Copy button and repeat-collapsing; panics land in the same
banner via a chained hook. Windows Chrome/Edge reporters can now paste
the exact ANGLE/D3D11 error instead of a screenshot of a blank canvas.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Every rfd dialog result funnels through sys::handle_path, which now
records the picked file's folder (persisted to the config dir), and all
pickers are built via sys::file_dialog(), seeded with that folder.
Save-As keeps the current tab's own directory as the first choice and
falls back to the remembered one.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Ctrl+V was globally bound to the entity paste command, so it never pasted
into an open text editor, and even when routed there iced's clipboard read
is a no-op on the web. Route Ctrl+V to the focused editor and, on the web,
read the system clipboard via the browser's async Clipboard API (the
keypress is a user gesture, so the read is permitted):
- MText editor: insert at the caret (web + native).
- TEXT field: append on the web; native lets the text_input paste itself.
- No editor open: falls through to the entity paste command as before.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Add an "OCS Web" button (bright ribbon-blue, desktop only) to the Start
page that opens the web build at hakanseven12.github.io/OpenCADStudio.
- Send Feedback now opens the GitHub issue with the body pre-filled:
app version + platform (OS/arch on desktop, browser user-agent on web),
via sys::platform_info + sys::percent_encode.
- Remove the redundant About button from the Start page (ABOUT command /
menu still opens it).
Native + web build; 75 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On the web, Save and Save As now open the in-app Save dialog (filename +
format) like the desktop, and the unsaved-changes prompt's Save routes
through it too — instead of an immediate rfd download. Confirm serializes
to bytes and downloads directly via a Blob + hidden `<a download>` click
(a user gesture), so there is no intermediate "click to download" link.
Replaces the rfd-based download_web / WebSaveResult round-trip with a
synchronous sys::download_bytes. Native save is unchanged.
Native + web build; 75 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Add DONATE to the Start-tab command allowlist; it was blocked there
with "No drawing open" while REPORT/CHANGELOG/ABOUT were allowed, so
the Donate button did nothing on the welcome screen.
- Implement sys::open_url on wasm via web_sys window.open(_blank) — it
was a no-op stub, so Donate, Release Notes and Send Feedback opened
nothing on the web. The click is a user gesture, so the pop-up blocker
allows it.
Native + web build.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
First step toward a web build: the library now compiles for
wasm32-unknown-unknown with --no-default-features.
- Add a `solid3d` feature (default on) gating truck-meshalgo /
truck-shapeops / lzma-sys. These pull vtkio → xz2 → lzma-sys (C),
which cannot cross-compile to wasm. Without the feature the 3-D
solid path (tessellation, booleans, ACIS import) becomes no-ops;
2-D CAD is unaffected. Native builds keep the feature by default.
- Move native-only crates (ureq, open) to non-wasm target deps and
add cfg shims: update-check is a no-op on wasm, and a new `sys`
module provides open_url + handle_path (rfd FileHandle → PathBuf;
synthesizes a name-based path on the web).
Native build + all 75 tests unchanged. Browser runtime wiring
(entry point, trunk, real file I/O) is a follow-up.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Karim Jerbi