`geom.tree().select()` silently returns zero results (or raises SWIG's
"An unknown error occurred") in the Linux release packages, while the same
commit built from source returns correct answers. `select_box()` agrees
between both, and geometry conversion is bit-identical -- only operations
that touch a stored TopoDS_Shape diverge.
Cause is linkage, not code. 8bdaa8c7c narrowed the Rocky builds from
`--shared` to `--ifcopenshell-shared`, which shares IfcOpenShell's own
libraries but leaves every dependency static. OCCT is then compiled
privately into each plug-in that uses it -- 15 of them, verified by their
own copies of the BRepClass3d/BRepExtrema/Standard_Failure strings.
That contradicts what the binaries already declare. tree.h:1748 casts a
`conversion_result_shape*` to `open_cascade_shape*`, moves the
TopoDS_Shape out of it and frees it; `open_cascade_shape` is defined once
in ifcopenshell_geometry_kernel_opencascade.so and left undefined in
ifcopenshell_geometry_tree_opencascade_brep.so for the loader to resolve.
So the two plug-ins are designed to share one OCCT-based type system, yet
static linking gives each its own Standard_Type registry and allocator.
Shapes get read and released by a different OCCT instance than made them.
Add `--opencascade-shared`, mirroring the existing `--ifcopenshell-shared`
precedent, and use it on both Rocky workflows. It cannot be spelled
`--occt-shared`: build-all.py parses any `occt-*` flag as a version
override.
BUILD_STATIC drives three things at once -- dependency link type,
-fvisibility=hidden, and BUILD_SHARED_LIBS -- so making one dependency
shared means overriding all three for it. Visibility is the subtle one:
OCCT's Standard_EXPORT expands to nothing on Unix, so it relies on default
visibility to export its API. Built shared under -fvisibility=hidden it
exports almost nothing and its own libraries cannot resolve against each
other (libTKMath.so fails to find
NCollection_BaseAllocator::CommonBaseAllocator in libTKernel.so). Static
archives are immune, which is why this surfaces only once OCCT goes
shared. Compile OCCT with the pre-visibility flag set instead.
Link the OCCT set with --as-needed. FindOpenCASCADE.cmake's config branch
uses OCCT's *complete* module list, Visualization included, which against a
static OCCT costs nothing -- an unreferenced module contributes no objects.
Against a shared OCCT all 47 become hard DT_NEEDED entries, and libTKV3d
pulls libGL.so.1 + libEGL.so.1, so `import ifcopenshell` fails on any
headless machine with "libEGL.so.1: cannot open shared object file" even
though nothing ever opens a window. Measured through the real find_package
path with the LINK_GROUP workaround composed: 67 DT_NEEDED without the
flag, 3 with it, TKV3d and TKOpenGl gone.
The flag is deliberately left open rather than closed with
-Wl,--no-as-needed. CMake emits the imported targets'
INTERFACE_LINK_LIBRARIES -- where OCCT lists libGL/libEGL -- after that
item, so closing the bracket switches the flag off immediately before the
libraries it exists to exclude. Verified against the shipped artifact:
closed, the kernel plug-in fell from 47 DT_NEEDED libTK entries to 14 and
lost TKV3d, yet still carried a direct libEGL.so.1 and still failed to
import on a headless server; open, the same 14 remain and libGL/libEGL are
gone. None of the 14 retained modules depends on GL.
Put the shared OCCT on LD_LIBRARY_PATH for the build itself. Nothing else
points at it -- IfcOpenShell's libraries get INSTALL_RPATH=$ORIGIN and OCCT
sits in its own dependency prefix -- so the post-build `import ifcopenshell`
check fails the same way. This is build-time only; the shipped packages get
libTK*.so* staged beside the payload with an $ORIGIN RUNPATH instead.
Suffix OCCT's install directory with `-shared` when it applies. Static and
shared installs are not interchangeable, but `build_dependency` skips any
dependency whose install dir already exists and cache_dependencies.py keys
its tarballs purely on that directory name -- so the static
`cache-occt-7.8.1.tar.gz` restored from the build-outputs repo silently
satisfied the build and BUILD_LIBRARY_TYPE was never applied. This is the
same cache stickiness 8bdaa8c7c described, pointing the other way. The
suffix makes the key configuration-aware, so it self-invalidates and the
static tarball stays valid for builds that still want static.
Packaging is the other half, and is why 8bdaa8c7c backed the flag out --
`stage_runtime_payload` only copies from install/ifcopenshell, so OCCT in
install/occt-* was never staged and `--shared` "worked by accident" off
cached static outputs. Stage libTK*.so* alongside, then give every staged
library an $ORIGIN RUNPATH: the core libs currently carry dead
build-machine RPATHs and the plug-ins carry none, resolving only because
the Python wrapper pulls them in by SONAME first. Shared OCCT has no such
first loader, since it is reached through the dlopen'd plug-ins.
The packages shrink: the python zip goes from 109.4 MB to 85.6 MB, because
the duplicated OCCT was 113 MB of the 287 MB unpacked payload (the eight
geometry_writer_ifc* plug-ins alone were 4.6 MB each) against ~67 MB for
one shared copy. Same argument as a91b1da28 ("Reduce Rocky package size")
and 402591e71.
macOS and Windows are affected too but are not fixed here. Their
packaging resolves via @loader_path install names and would need
install_name_tool rewriting, which cannot be verified from Linux; adding
the flag without that would ship a package that fails to load.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
As we're not currently bundling dlls for all other external dependencies.
It seems `--shared`in CI previously worked sort of by accident - since it was relying on the cached build outputs that were previously built statically.
This branch carried v0.8.0's strict `[tool.ty.rules] all = "error"` config but
not the source fixes that were made upstream to satisfy it, so both ci-lint ty
gates were failing: `poe ty-ios` reported 256 diagnostics and `poe ty-bonsai`
258. Both are now clean.
Most fixes are ported from v0.8.0 and follow two idioms: initialise a name
before a conditional that may not bind it (plus an `assert` where the invariant
is real but not provable), and close an exhaustive `if`/`elif` chain with
`else: assert False, <discriminant>`.
The branch's own newer accessors are preserved throughout - `.file`,
`.declaration`, `file.types()`, `get_max_id()` are kept rather than reverted to
`wrapped_data.*`, and non-ty upstream changes (notably the in-progress geometry
cache removal) are deliberately not pulled in.
Notable fixes that are not straight ports:
* ifcopenshell_wrapper.pyi: `entity_instance.file` was declared as
`def file(self) -> file`, where the property name shadows the `class file`
below it, so the annotation resolved to `Unknown`. Every `element.file` in
the codebase was therefore unchecked. Qualifying it to `ifcopenshell.file`
restores `.schema` to its Literal union and surfaces no new diagnostics.
* model/wall.py: a duplicated merge fragment in the void-straddle path ran an
always-true `if void_straddles:` that read `new_opening` from the mutually
exclusive branch (stale value, or NameError on the first iteration), followed
by an unreachable duplicate `elif`. Removing it makes the file match v0.8.0.
* light/operator.py: upstream's own fix unpacks three targets from two values
and raises ValueError unconditionally; corrected to `None, None, None`.
* assign_system.py, validate.py, geom/main.py: walrus-in-genexp is valid at
runtime (PEP 572 binds in the containing scope) but ty does not model it;
rewritten as explicit loops, matching upstream.
Verified: poe ty-ios, poe ty-bonsai, ruff check src/ nix/, black --check .,
and compileall -W error at py3.10 (ifcopenshell-python) and py3.11 (bonsai).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Last reference to this file was dropped in 7ae685dbf, though the ref was
pointing to `/patches/opencollada/pr622.patch`, so IIUC
`patches/pr622.patch` was never used.
(cherry picked from commit c013b9aca7)
Introduced in e21277e80, reference removed
in 683cadeb7 when occt was bumped to 7.3.0 and switched to git-tag based
download.
(cherry picked from commit 24e454ce0c)
Error was:
```
configure: error: could not find a working compiler, see config.log for details
```
config.log:
```
conftest.c: In function 'f':
conftest.c:12:48: error: too many arguments to function 'g'; expected 0, have 6
12 | for(i=0;i<1;i++){if(e(got,got,9,d[i].n)==0)h();g(i,d[i].src,d[i].n,got,d[i].want,9);if(d[i].n)h();}}
| ^ ~
```
d390911d75 ("build_osx: build BonsaiViewer on macOS via build-all.py",
2026-06-01) accidentally committed a local single-version pin
(`PYTHON_VERSIONS = ["3.11.8"]`) intended only for fast iteration
during macOS bring-up. With macOS / CI green and the Python wrapper
fix from 748b4e72a landed, restore the full multi-version list so
both Rocky and macOS CI publish wrappers for 3.10/3.11/3.12/3.13/3.14
again.
Cost is ~5 from-source Python builds per CI run; the cache-deps
plumbing in nix/cache_dependencies.py already memoises these so
repeated runs only pay it once per Python release bump.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
# Why this exists
The bonsai macOS CI (`build_osx.yml`, arm64) currently fails the
`IfcOpenShell-Python` smoke test with:
ImportError: dlopen(.../_ifcopenshell_wrapper.cpython-311-darwin.so,
0x0002):
Library not loaded: @rpath/ifcopenshell.document.rdb.dylib
Reason: tried: '$ORIGIN/ifcopenshell.document.rdb.dylib'
(no such file)
`_ifcopenshell_wrapper.cpython-311-darwin.so` has a hard `LC_LOAD_DYLIB`
of `@rpath/ifcopenshell.document.rdb.dylib` and its only `LC_RPATH` is
`$ORIGIN` (= `site-packages/ifcopenshell/`). The plug-in dylib is not
present at that path on macOS, so the wrapper fails to load and the
build smoke test (`build-all.py: compile_python_wrapper`) errors out.
BonsaiViewer.app builds, installs, and macdeployqt-deploys cleanly
before this point — the failure is downstream and unrelated to wgpu,
BonsaiViewer, or anything else on this branch.
# Where the regression came from
Two commits on the branch line that became `ifcviewer-wgpu`:
b599ee10 "More work on isolating into plug-ins" (2026-04-18, Thomas Krijnen)
b022ca7e7 "Some plug-in work" (2026-04-21, Thomas Krijnen)
`b599ee10` added a hard link dep:
target_link_libraries(ifcopenshell_wrapper PRIVATE document_serializer_rdb)
which bakes `@rpath/ifcopenshell.document.rdb.dylib` into the wrapper's
`LC_LOAD_DYLIB`. `b022ca7e7` added a `if(CREATE_BUNDLE) ...
install(TARGETS ${_ifcopenshell_python_runtime_targets}
LIBRARY DESTINATION "${python_package_dir}/ifcopenshell" ...)` block
that was *intended* to satisfy that link dep by copying plug-ins next
to the wrapper. On macOS arm64 the install rule does not actually
deposit `ifcopenshell.document.rdb.dylib` into
`site-packages/ifcopenshell/`, so the runtime dlopen fails.
# Why 227d85d worked
`227d85d` (2026-05-15) is on the `v0.8.0` line, not on the
`datamodel-v1.0 -> ifcviewer -> ifcviewer-wgpu` line. The merge-base
of `227d85d` and `ifcviewer-wgpu` is `e6258ab4` (2026-04-13). Both
b599ee10 and b022ca7e7 live on the wgpu side of that fork and are not
ancestors of `227d85d`:
$ git merge-base --is-ancestor b599ee10227d85d
[exit 1 — NOT an ancestor]
$ git merge-base --is-ancestor b599ee10 v0.8.0
[exit 1 — NOT an ancestor]
So the macOS Python wheel built fine on `v0.8.0` because that branch
never had the plug-in refactor; it has been broken on our branch line
since 2026-04-21. Nobody noticed because nobody had been firing
`build_osx.yml` against this branch line until this week's bonsai CI
work.
# What this commit does
Adds an `IFCOS_BUILD_PYTHON_WRAPPER` env var to `nix/build-all.py`.
Defaulting to `on` preserves existing behaviour everywhere; setting
it to `off` (or `0`/`false`/`no`) drops `IfcOpenShell-Python` from
the target set so `build-all.py` skips the wrapper build + smoke
test entirely.
`build_osx.yml` sets `IFCOS_BUILD_PYTHON_WRAPPER=off` so the bonsai
macOS CI can complete and upload `BonsaiViewer.app` while the plug-in
install rule is broken.
# What Thomas should do
Once the install rule in `src/ifcwrap/CMakeLists.txt` (the
`if(CREATE_BUNDLE) ... install(TARGETS ${_ifcopenshell_python_runtime_targets}
LIBRARY DESTINATION "${python_package_dir}/ifcopenshell" ...)` block,
added in b022ca7e7) is fixed to actually drop
`ifcopenshell.document.rdb.dylib` next to the wrapper in
site-packages on macOS — this commit can be reverted in its entirety:
the env-gate in `build-all.py` AND the `IFCOS_BUILD_PYTHON_WRAPPER=off`
in `build_osx.yml`. The bonsai macOS workflow will then build the
Python wrapper too.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Linux (build_rocky.yml) and Windows (win/build-all-win.py) already pass
BUILD_BONSAIVIEWER=ON when building from source; macOS was the odd one
out. Three small changes to bring it up to parity:
1. .github/workflows/build_osx.yml — \`brew install qt\` (Qt6 with Svg)
in the Install Dependencies step, then set QT_DIR=\$(brew --prefix
qt) and BUILD_BONSAIVIEWER=ON in the Run Build Script env.
2. nix/build-all.py — install_qt6() now honours a pre-set QT_DIR.
Before this change get_qt6_aqt_config() raised on non-Linux,
blocking bonsai builds on macOS/Windows from ever using a
system-provided Qt6. We now validate that QT_DIR points at a real
Qt6 install (probes lib/cmake/Qt6/Qt6Config.cmake) and skip the aqt
download path if so. Linux flow is unchanged: when QT_DIR is unset
the function falls through to the existing aqtinstall path.
build_osx.yml stays workflow_dispatch-only — slow run (~1h with
ccache, longer cold), so manual fire when wanted. Cherry-pick this
file + nix/build-all.py to v0.8.0 to make the workflow dispatchable
against the ifcviewer-wgpu branch before the squash lands.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Compile the Bonsai Viewer as part of the Linux and Windows binary builds,
and ship the Autodesk connector alongside the viewer executable.
Qt6 dependencies:
- The viewer links Qt6::Svg for runtime icon tinting. Svg is a separate
base-Qt archive, so aqt now installs "qtbase qtsvg" (plus icu on Linux)
rather than qtbase alone, on both Linux and Windows.
- Qt6::CorePrivate is exposed differently across Qt versions: Qt 6.8 ships
the target inside Qt6Core, while Qt 6.10 provides it only as a separate
CorePrivate config package. The viewer CMakeLists requests it via
OPTIONAL_COMPONENTS so it resolves on both.
- When cross-compiling Windows ARM64, windeployqt runs from the host x64
Qt, so qtsvg is installed into the host Qt as well.
Windows build:
- build-all-win.py passed -DBUILD_IFCVIEWER, a flag since renamed to
BUILD_BONSAIVIEWER, so the Windows build compiled no viewer at all. It
now passes -DBUILD_BONSAIVIEWER.
- The Autodesk connector is bundled under connectors/ next to
BonsaiViewer.exe in the packaged archive, mirroring the Linux builds.
- The Windows workflow builds the connector (PyInstaller) before the main
build so it is available to bundle.
Connector bundling:
- The Linux rocky workflows build the connector and bundle it into the
BonsaiViewer archive; the Windows build now does the same.
Generated with the assistance of an AI coding tool.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replace the old IFC viewer build switch with BUILD_BONSAIVIEWER in CMake, the Linux workflows, and the nix build script.
Generated with the assistance of an AI coding tool.
Dion Moult