subagent transcript

session 17f09f89 · workflow wf_1ced79bb

delegated task

FreeCAD is ported to WebAssembly (JSPI + native-EH, static CPython 3.14, no dlopen, no subprocess, single-thread, Qt 6.11). 17 workbenches already wor

▸ Magik

FreeCAD is ported to WebAssembly (JSPI + native-EH, static CPython 3.14, no dlopen, no subprocess, single-thread, Qt 6.11). 17 workbenches already work incl. Part/PartDesign/Sketcher/Measure/Draft/Assembly/TechDraw/Mesh + numpy + PySide6 + pivy._coin. Source tree: /home/magik6k/lcad-wasm/freecad-port/FreeCAD (branch wasm-parity-push). Build tree (already configured, DO NOT build — read only): /opt/toolchains/src/freecad-gui-build. The module-enablement pipeline is documented in the memory files at /home/magik6k/.claude/projects/-home-magik6k-lcad-wasm/memory/ (read freecad-wasm-parity-roadmap.md and freecad-wasm-viewport.md FIRST — they describe the src/Main/CMakeLists.txt inittab generator, FREECAD_WASM_EXTRA_PY_MODULES / FREECAD_WASM_EXTRA_LINK_LIBS, SHARED->STATIC conversion, data-packaging with file_packager, and known gotchas). Constraints that shape feasibility: NO subprocess/fork/exec, NO dlopen (all C-ext modules static via inittab), NO threads (QThreadPool/QtConcurrent absent), NO listen sockets, wasm32 4GB ceiling. C++ deps must be new-EH (-fwasm-exceptions -sWASM_LEGACY_EXCEPTIONS=0). Pure-Python addons are fine; native ones need static linking. Your job: READ ONLY. Do not edit source, do not run cmake/ninja/emcc. Produce an actionable plan a build engineer can execute. Here are 5 parallel parity investigations (JSON): [{"area":"CAM (Path) workbench","verdict":"tractable-with-work","effortHours":"8-16","valueToUsers":"high","rootBlockers":["pybind11 is not installed anywhere in the toolchain (no headers under /opt/toolchains, not in the venv, FREECAD_USE_PYBIND11=OFF). BUILD_CAM auto-forces FREECAD_USE_PYBIND11=ON which calls find_package(pybind11 REQUIRED) -> hard configure failure until pybind11 is provided. This is also the FIRST pybind11 module in the wasm build, so pybind11's exception-translation + shared-internals under JSPI + native wasm-EH is unproven (no fundamental reason it should fail; header-only, RTTI+exceptions available).","src/Main/CMakeLists.txt has NO BUILD_CAM block adding the modules to the inittab/link lists (only a stubbed Path->PathApp name mapping exists at lines 116-118). Without it the static PathApp/area/tsp_solver/PathGui/PathSimulator archives are never linked or registered.","The 'packaging' Python module is imported UNGUARDED at CAM workbench init (InitGui.py:109 'from packaging.version import Version, parse' and GuiInit.py:101) and is not confirmed present in any deploy .data package; workbench Initialize() will raise ImportError and fail to load unless packaging (pure-Python, present in /opt/toolchains/venv/lib/python3.14/site-packages/packaging) is added to the wasm FS.","The full CAM Python tree (~400 .py under src/Mod/CAM: Path/**, PathScripts, Machine, Tools) is not in the base freecad.data and must be FS-packaged via a supplementary file_packager .data (pack-mods.sh), deduped against existing packages."],"concreteSteps":["Provide pybind11 to CMake: `pip install pybind11` into /opt/toolchains/venv (header-only, arch-independent, safe for the emscripten cross-build). At reconfigure pass `-Dpybind11_DIR=$(/opt/toolchains/venv/bin/python -m pybind11 --cmakedir)`. pybind11 headers only need Python.h (already the wasm CPython include already used by the CAM targets, which link Python3::Python directly and pass ${pybind11_INCLUDE_DIR} manually) so find_package resolving host-Python is harmless here.","Reconfigure the existing build tree (do NOT re-run emcmake): `cmake -S /home/magik6k/lcad-wasm/freecad-port/FreeCAD -B /opt/toolchains/src/freecad-gui-build -DBUILD_CAM=ON -Dpybind11_DIR=...`. REQUIRES_MODS(BUILD_CAM BUILD_PART BUILD_MESH) is already satisfied (both ON). This auto-turns FREECAD_USE_PYBIND11 ON (InitializeFreeCADBuildOptions.cmake:230).","Edit src/Main/CMakeLists.txt: add a BUILD_CAM block after the BUILD_ASSEMBLY block (~line 79) registering the five static modules. The generator already maps Path->PathApp; area/tsp_solver/PathSimulator/PathGui are plain names matching their PyInit_* symbols (verified: PYBIND11_MODULE(area)/(tsp_solver), PyMOD_INIT_FUNC(PathSimulator)/(PathGui)). No SHARED->STATIC edits needed anywhere (global add_library override at CMakeLists.txt:96 handles it).","FS-package the CAM Python tree + the `packaging` dependency: run /home/magik6k/lcad-wasm/pack-mods.sh CAM to build deploy/freecad-mods-cam.data(.js/.gz), add packaging/ (copy /opt/toolchains/venv/lib/python3.14/site-packages/packaging into the package or the python stdlib zip), and load it in index.html before FreeCAD.js. Dedup against base freecad.data to avoid the opaque createFreeCAD '[object Object]' duplicate-path rejection.","Build: `cd /opt/toolchains/src/freecad-gui-build && ninja FreeCAD.js`. Expect the heaviest new TUs to be libarea/clipper.cpp, libarea/Adaptive.cpp, App/Area.cpp (140KB) and the two pybind11 TUs (memory-heavy). area-native transitively pulls the already-built Clipper2::Clipper2Z and Import.","Deploy + runtime-verify: copy bin/FreeCAD.{js,wasm} to deploy, regen .gz, load the CAM .data. Probe in-browser: `import PathApp, area, tsp_solver, PathSimulator`; `Gui.activateWorkbench('CAM_Workbench')`; run a Pocket/Profile op (exercises area.makeSections + Adaptive2d) and tsp_solver.solve to confirm the pybind11 path works under JSPI.","Leave the OCL-dependent ops (Surface/Waterline/RotarySurface, import ocl) degraded — they are already internally try/except-guarded (Surface.py:36-41) and not imported at workbench init, so they don't block boot; opencamlib is a separate unported native module out of scope."],"risks":"\"Primary risk is pybind11 being the first pybind11 module in this JSPI+new-EH build: its C++ exception translation (py::cast_error/error_already_set caught in tsp_solver_pybind.cpp) and per-process 'internals' capsule shared between the two pybind11 modules (area, tsp_solver) must work with -fwasm-exceptions/-sWASM_LEGACY_EXCEPTIONS=0; both compiled by the same toolchain so ABI-id matches, and no dlopen is involved, but …

Fable

thinking

I'm working through the synthesis of five parallel parity investigations into a single actionable plan, ranking them by value and effort to produce the structured output. The investigations are already detailed, so my focus is on evaluating and organizing them effectively.

Looking at the five targets: the persistence reload fix stands out as the highest value-to-effort ratio at just 1-2 hours for a localized C++ edit and rebuild. The BIM/Arch workbench comes next at 3-6 hours with high value and no native relink needed—just a host-side reconfigure, one guard edit, and packaging. CAM is substantial at 8-16 hours but blocked on pybind11 installation and CMakeLists configuration. Plot and matp  …

Fable

I'll verify a few key claims for my top-ranked targets before synthesizing.

Fable

⚙ Bash

{"command": "cd /home/magik6k/lcad-wasm/freecad-port/FreeCAD && sed -n '1580,1600p;1900,1930p' src/Base/Parameter.cpp 2>/dev/null; echo \"---HEADER---\"; grep -n \"_Reset\\|_RebindGroups\\|_GroupMap\" src/Base/Parameter.h", "description": "…

tool result

    }
}

void ParameterGrp::_Reset()
{
    _pGroupNode = nullptr;
    for (auto& v : _GroupMap) {
        v.second->_Reset();
    }
}

//**************************************************************************
//**************************************************************************
// ParameterSerializer
//++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
ParameterSerializer::ParameterSerializer(std::string fn)
    : filename(std::move(fn))
{}

P  …[truncated]

Fable

⚙ Bash

{"command": "cd /home/magik6k/lcad-wasm/freecad-port/FreeCAD && sed -n '70,120p' src/Main/CMakeLists.txt", "description": "Verify Main CMakeLists module registration block"}

tool result

        list(APPEND FREECAD_WASM_PY_MODULES DraftUtils)
    endif()
    if(BUILD_TECHDRAW)
        list(APPEND FREECAD_WASM_PY_MODULES TechDraw)
        list(APPEND FREECAD_WASM_PY_GUI_MODULES TechDrawGui)
    endif()
    if(BUILD_ASSEMBLY)
        list(APPEND FREECAD_WASM_PY_MODULES Assembly)
        list(APPEND FREECAD_WASM_PY_GUI_MODULES AssemblyGui)
    endif()
    if(BUILD_START)
        list(APPEND FREECAD_WASM_PY_GUI_MODULES StartGui)
    endif()
    if(NOT BUILD_GUI)
        set(FREECAD_  …[truncated]

Fable

⚙ Bash

{"command": "cd /home/magik6k/lcad-wasm/freecad-port/FreeCAD && sed -n '20,60p' src/Mod/BIM/nativeifc/ifc_openshell.py", "description": "Verify BIM ifc_openshell eager imports"}

tool result

# *   License along with FreeCAD. If not, see                               *
# *   <https://www.gnu.org/licenses/>.                                      *
# *                                                                         *
# ***************************************************************************

"""Utilities to help people verify and update their version of ifcopenshell"""

from packaging.version import Version

import FreeCAD
import FreeCADGui
from addonmanager_utilities import   …[truncated]

Fable

thinking

I've verified the key locations across the codebase: the persistence fix with `_Reset()` ending at line 1589 and LoadDocument returning at 1927, the header declarations at 439 and 490, the BIM eager imports at lines 27 and 31, and the CMakeLists module registration around line 76-78 with the Path-to-PathApp mapping. I have what I need to put this together.

Now I'm ranking the fixes by effort and impact. The persistence reload fix in Parameter.cpp is the clear winner—just 1-2 hours of work, high value, a single localized rebuild with minimal risk. BIM and Arch support come next at 3-6 hours with no wasm relink needed, then CAM/Path at 8-16 hours as the first real pybind11 challenge. There ar  …

Fable

All key claims verified: `_Reset()` ends at Parameter.cpp:1589, `LoadDocument` returns at 1927, header has `_Reset()` at 439 and `_GroupMap` at 490; BIM eager imports at ifc_openshell.py:27 and :31; Main CMakeLists module-registration block and Path→PathApp mapping confirmed. Synthesizing the plan.

Fable

⚙ StructuredOutput

{"rankedTargets": [{"rank": 1, "target": "DEFERRED persistence reload fix — repopulate live ParameterManager on reload (ParameterGrp::_RebindGroups)", "why": "Highest value/effort ratio of everything here: ~1-2h, a single localized C++ chan…

tool result

Structured output provided successfully