os: add support for windows by AmyJeanes · Pull Request #2427 · commaai/panda

AmyJeanes · 2026-09-07T20:54:03Z

Native Windows support for panda's firmware build and libpanda test library (MSYS2 CLANG64, clang), plus a windows-latest CI entry. openpilot's Windows build uses this to build and flash panda.

SCons uses the mingw tool with clang on Windows (the default picks MSVC); the .dll target and cffi path are named per platform.
CANPacket_t keeps its wire layout: the DLL builds with -mno-ms-bitfields and the flag bitfields are declared unsigned int so cffi's MSVC rules match the firmware. Linux/macOS layout unchanged (same fix as opendbc's libsafety).
The signing step runs sign.py through the SCons interpreter (cmd.exe honours neither a shebang nor a VAR=value prefix).
The dev toolchain and cppcheck come from their PyPI wheels; the setup and test scripts use the venv's python.

Line endings were split out and comma has already merged them.

Draft: the one TEMP commit at the tip (marked TODO REMOVE) drops once commaai/dependencies#107 publishes the win_amd64 toolchain/cppcheck wheels; it points uv at a pre-release index meanwhile.

Part of a six-repo series: commaai/openpilot#38810, commaai/msgq#709, commaai/opendbc#3724, commaai/rednose#61, commaai/dependencies#107. dependencies merges first; the rest independently.

Generated with Claude, human-reviewed and tested locally and in CI.

--- TODO REMOVE AFTER THE SUBMODULE PRS MERGE (commaai/msgq#709, commaai/panda#2427, commaai/rednose#61, commaai/opendbc#3724) --- Until the four submodule PRs merge, the pointers move to the windows-tidy branches on my fork and .gitmodules fetches them from there: panda and opendbc just below their own TEMP index commits (their uv sources would leak into this lockfile), rednose at its TEMP commit (its source supplies comma-deps-eigen for Windows), msgq's series on openpilot's current pin (windows-tidy-pin). The lock follows the submodules' metadata. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

Split from #2427.

SCons' default tools pick MSVC there; the mingw tool with clang builds both. CANPacket_t is copied onto the wire, so libpanda needs GCC's packed bitfield layout (-mno-ms-bitfields) and cffi's MSVC rules must produce the same one: unsigned int flag bits, like addr, do on both. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

`SETLEN=1 sign.py ...` is a shell command cmd.exe cannot run. The interpreter running SCons has the dependencies, and SETLEN moves into the command's environment; the same command on every platform. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

The release shim branches only know Linux and macOS; the comma-deps wheels cover Windows too (import names unchanged). They declare Python >=3.12, so the marker keeps the 3.11 half of requires-python resolving. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

An MSYS2 CLANG64 shell, like openpilot's Windows job; setup.sh installs clang, git and uv there with pacman. The misra script calls python (a venv on Windows has no python3), and the mutation test runs it through the bash on PATH, since cmd.exe cannot. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

--- TODO REMOVE AFTER THE SUBMODULE PRS MERGE (commaai/msgq#709, commaai/panda#2427, commaai/rednose#61, commaai/opendbc#3724) --- Until the four submodule PRs merge, the pointers move to the windows-tidy branches on my fork and .gitmodules fetches them from there: panda and opendbc just below their own TEMP index commits (their uv sources would leak into this lockfile), rednose at its TEMP commit (its source supplies comma-deps-eigen for Windows), msgq at its tip. The lock follows the submodules' metadata. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

Quote $DIR (and the misra script's paths) in the setup and test scripts so a checkout path containing spaces works. Cross-platform: unquoted, these break on Linux too, spaces are just far more common on Windows. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

--- TODO REMOVE AFTER THE SUBMODULE PRS MERGE (commaai/msgq#709, commaai/panda#2427, commaai/rednose#61, commaai/opendbc#3724) --- Until the four submodule PRs merge, the pointers move to the windows-tidy branches on my fork and .gitmodules fetches them from there: panda and opendbc just below their own TEMP index commits (their uv sources would leak into this lockfile), rednose at its TEMP commit (its source supplies comma-deps-eigen for Windows), msgq at its tip. The lock follows the submodules' metadata. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

debug_console.py polled stdin with select(), which on Windows only works on sockets, so som_debug.sh failed there. Poll the console with msvcrt on Windows and keep select() elsewhere. Also quote `cd "$DIR"` in som_debug.sh so a checkout path with spaces works. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

…ies#107 publishes --- TODO REMOVE AFTER DEPENDENCY PR MERGES (commaai/dependencies#107) --- Dropped once the win_amd64 wheels are on PyPI; until then uv takes comma-deps-gcc-arm-none-eabi and comma-deps-cppcheck for win32 from a release on my fork built from that PR. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AgqYZYWLpwE2T5vS3nVEt

AmyJeanes changed the title ~~os: add support for windows~~ os: add support for windows (1/2) Sep 7, 2026

AmyJeanes mentioned this pull request Sep 7, 2026

os: add support for windows commaai/dependencies#107

Open

AmyJeanes marked this pull request as draft September 7, 2026 23:22

AmyJeanes changed the title ~~os: add support for windows (1/2)~~ os: add support for windows Sep 7, 2026

AmyJeanes mentioned this pull request Sep 8, 2026

os: add support for windows commaai/openpilot#38810

Draft

adeebshihadeh mentioned this pull request Sep 9, 2026

Use LF line endings #2428

Merged

adeebshihadeh pushed a commit that referenced this pull request Sep 9, 2026

Use LF line endings

35621ee

Split from #2427.

AmyJeanes and others added 4 commits September 9, 2026 18:14

AmyJeanes force-pushed the windows branch from 0c5376b to 63107cc Compare September 9, 2026 21:05

AmyJeanes force-pushed the windows branch from 63107cc to 49a42d6 Compare September 10, 2026 11:17

AmyJeanes force-pushed the windows branch from 49a42d6 to 8c5746e Compare September 10, 2026 14:47

Navigation Menu

Uh oh!

os: add support for windows - #2427

os: add support for windows#2427
AmyJeanes wants to merge 7 commits into
commaai:masterfrom
AmyJeanes:windows

AmyJeanes commented Sep 7, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

AmyJeanes commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

AmyJeanes commented Sep 7, 2026 •

edited

Loading