Stabilize new build in separate repository, and get CI/CD working - #8
Open
CodeByDrescher wants to merge 54 commits into
Open
Stabilize new build in separate repository, and get CI/CD working#8CodeByDrescher wants to merge 54 commits into
CodeByDrescher wants to merge 54 commits into
Conversation
Collaborator
Author
CodeByDrescher
force-pushed
the
stabilize-new-build
branch
17 times, most recently
from
July 10, 2026 14:52
2d09965 to
19e878b
Compare
Also: adding windows on arm (woa) support(?)
Both libraries are shared with vcell-stochastic, which already consumes them as submodules. Vendoring them here let the copies drift apart. vcell-messaging: the public API is byte-identical to the copy it replaces (the apparent diff was CRLF vs LF), so no call sites change. The submodule additionally carries a deadlock fix and a shutdown fix: the lock order is now queue-before-stop rather than the reverse, and shutdown joins the worker instead of waiting for the queue to drain -- an empty queue is not "all work finished", since the worker pops under the lock but sends after releasing it. vcell-expressionparser: the in-tree copy here was the *newer* of the two -- the std::format/nullptr/warning cleanup from f3c3107 and 096ba48 existed only in this repo. Rather than lose it, that work was pushed up to the submodule (virtualcell/vcell-expressionparser @ 2fd45ae) and is pinned here, so the two repos converge on the better copy instead of the older one. The target rename ExpressionParser -> vcellexpressionparser is reflected in IDAWin. ExpressionParserTest/ is left alone; it was already wired into nothing. Verified: clean configure, full build, and all 4 gtest cases pass. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
OdeResultSet.cpp and VCellSundialsSolver.cpp call memset/memcpy without including <cstring>. Older libstdc++ pulled it in transitively through other headers; GCC 13 no longer does, so both fail to compile with "'memset' was not declared in this scope". Pre-existing and unrelated to the submodule migration -- it reproduces on a pristine checkout of this branch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Added support for recursive submodule checkout in CI workflow.
The windows-latest runner moved to the windows-2025-vs2026 image, which
ships only Visual Studio 18 (MSVC 14.51). The x86_64 Conan profile pinned
compiler.runtime_version=v143, and for compiler=clang Conan derives the
Visual Studio version from that setting rather than from
tools.microsoft.msbuild:vs_version (see _vcvars_versions() in
conan/tools/microsoft/visual.py, which maps v143/v144 -> VS 17 and
v145 -> VS 18). VCVars therefore looked for a VS 17 install that no longer
exists and the job failed while building fmt from source:
ConanException: VS non-existing installation: Visual Studio 17
Bump the x86_64 profile to v145. Windows-ARM64 stays on v144 because the
windows-11-arm64 image still carries VS 17; it will need the same bump when
that image rolls forward.
Also record the runtime_version -> Visual Studio mapping in CLAUDE.md and in
the profile, since the setting's name does not suggest it selects a Visual
Studio installation.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Five issues introduced by this branch and left behind when it replaced the pybind11/scikit-build scaffold with ctypes + uv_build: - version() was annotated `-> str` but returned the ReturnValue pydantic model from call_version() unchanged, so callers got `success=True message='...'` instead of the version string. Unwrap `.message`, matching how solve() already unwraps `.success`. The existing test only asserted `is not None`, so it caught nothing either way. - .gitignore listed __pycache__ per directory and missed pyvcell_odesolver/_internal/, which appeared as untracked .pyc files. Replace the three entries with a single __pycache__/ rule. - [tool.ruff] src pointed at "src", a directory that has never existed in this layout (the package sits at the repo root), leaving first-party import resolution misconfigured. - argparse was fetched via FetchContent with no GIT_TAG, so every clean configure cloned whatever was on the default branch. Pin v3.2 to match the ">=3.2 <4.0" range conanfile.py already declares; googletest alongside it was already pinned. - extern/pybind11 (3.4 MB) had no remaining references anywhere -- no add_subdirectory, no find_package -- once the bindings became ctypes. CLAUDE.md documented the first four as known defects; updated accordingly. Verified: cmake configure/build clean with argparse pinned at v3.2, ctest 4/4, pytest 2/2, and version() now returns a str. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Stabilize the ODE solver against sanitizers, modernize the build, and ship a
Python distribution alongside the native CLI.
argparse,spdlog,libcurl); CMake minimum 3.13 → 3.16; consolidate platform detection onWIN32; new optionsOPTION_TARGET_PYTHON_BINDING,OPTION_TEST_WITH_LOCALHOST,OPTION_STATICALLY_LINK; warning cleanuppass and
sprintfremoval so the C++20 build compiles clean.VCellSolverInput+VCellSolverFactory; introduce abstractVCellSolverbase; replace the 273-line hand-rolled argv parser in
SundialsSolverStandalonewithargparse; expose a reusablesolve(input, output, tid)entry inSundialsSolverInterface(used byboth the CLI and the Python module).
SimulationMessagingintoMessageEventManager(worker thread + queue, with documented lockordering),
CurlProxyClasses(AbstractCurlProxy/NullCurlProxy/CurlProxy— replaces#ifdef USE_MESSAGINGforests with polymorphism),JobEventStatus(now aJobEvent::Statusnamespaced enum), andWorkerEvent. Fixes from TSAN/UBSAN/leak runs: data race,mutex-access-before-constructor-finished, undefined behavior, memory
leaks, int overflow in a test.
std::jthread→std::threadforcompiler portability. Net:
SimulationMessaging.cpp755 → 194 lines.pyvcell_odesolver): scikit-build-core + vendoredpybind11;src/main.cppexposesversion()andsolve(cvode_input_file_path, output_file_path, tid=-1).arm64, Python 3.10–3.13. Per-platform Conan profiles,
cibuildwheelPython wheels,
lipomacOS Universal build, GHCR manylinux images,moldlinker on Linux, sanitizer-enabled CMake.IDAWin/tests/smoke/smoke.pywith a top-leveltests/dir of GoogleTest C++ tests (smoke, hello, message-processing)plus a
test_basic.pythat exercises the wheel.Test plan
ctest -VVgreen locally (Linux, macOS); Windows via CISundialsSolverStandalone_x64runs the smoke.cvodeInputandmatches
*.ida.expected/*.cvode.expectedpip installof the built wheel exposespyvcell_odesolver.solve(...)and produces output matching the CLI-DOPTION_TARGET_MESSAGING=ONagainst a localhost broker(
-DOPTION_TEST_WITH_LOCALHOST=ON) — no TSAN/leak warningslipo -archs