[ErenAta16]

This PR was authored with AI assistance (Claude Code); I reviewed every changed line and take responsibility for it.

Summary

Fixes #23989

On Windows the wheel packages flatc through setup.py's get_executable_name(), so the binary is executorch/data/bin/flatc.exe. _get_flatc_path() in exir/_serialize/_flatbuffer.py asked executorch.data.bin for flatc, is_file() was False, and the lookup fell back to a bare flatc on PATH. That resolves only when the environment's Scripts directory is on PATH (pip puts the flatc console-script shim there). Called by the interpreter's full path, to_executorch() after XnnpackPartitioner raised FileNotFoundError: [WinError 2].

The change asks for flatc.exe on win32 and leaves the other platforms and the first candidate (flatbuffers-flatc, used by the Buck/PAR build) as they were. This completes #19899, which added the executorch.data.bin candidate with the Linux/macOS spelling only.

All eight modules that run flatc import the same helpers from this file, so no other call site needs a change; the audit is in #23989.

Test plan

Added TestGetFlatcPath to exir/_serialize/test/test_flatbuffer.py. It fakes an executorch.data.bin directory containing only the wheel's filename and checks that _get_flatc_path() returns that file: flatc.exe with sys.platform set to win32, flatc for linux and darwin. It patches the module's cache globals so it does not depend on test order.

python -m pytest exir/_serialize/test/test_flatbuffer.py -k GetFlatcPath
python -m pytest backends/xnnpack/test/serialization

Both run on the Windows job (pytest-windows.ini lists exir/_serialize/test and backends/xnnpack/test/serialization).

Measured on Windows 11 x64, Python 3.12.10, a fresh venv with the PyPI executorch==1.5.1 win_amd64 wheel and torch==2.14.0+cpu, FLATC_EXECUTABLE unset, no flatc on PATH, interpreter called by full path. I did not build the wheel from source: for "after" I replaced exir/_serialize/_flatbuffer.py inside that install with the file from this branch. The _get_flatc_path() block in the 1.5.1 wheel is identical to main, so the "before" is the same code.

                                                      before                after
TestGetFlatcPath (new)                                1 failed, 1 passed    2 passed
test_serialize_xnnpack_binary (existing)              FileNotFoundError     passed
                                                      [WinError 2]
exir/_serialize/test + xnnpack serialization,
  failing test ids                                    6                     4
minimal Linear, to_edge_transform_and_lower +
  XnnpackPartitioner + to_executorch                  WinError 2            OK, .pte runs and matches eager

The failing set after the change is a strict subset of the set before it: the two that went green are the new test and test_serialize_xnnpack_binary, and nothing new fails. The 4 that remain (test_named_data_store x2, test_program x1, test_ptn x1, the last with several failing subtests) fail identically with and without this change. The first three are assertions about named-data alignment and segment handling whose fixes (#23517, #23340) landed on main after v1.5.1, so I believe they are main's tests running against the older 1.5.1 wheel code; I did not run them against a build from main, and I did not trace the test_ptn one.

Not covered: I could not run the new test on Linux or macOS (only its linux/darwin cases with sys.platform patched on Windows), and nothing was run under Buck. ufmt 2.8.0 / usort 1.0.8 / black 24.4.2 and flake8 6.1.0 at the versions in requirements-lintrunner.txt are clean on both files; I did not run the full lintrunner (no mypy/pyre).

cc @nil-is-all

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/23992

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⚠️ 1 Awaiting Approval

As of commit 09a34c5 with merge base adc713a ():

AWAITING APPROVAL - The following workflow needs approval before CI can run:

• periodic (gh)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: ErenAta16 / name: ErenAta16 (09a34c5)

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.