Skip to content

Promote …f8 to a known input characteristic (issue #15 field report) - #22

Merged
mlstr0m merged 9 commits into
mainfrom
agent/funny-dirac-vjofbi
Sep 28, 2026
Merged

Promote …f8 to a known input characteristic (issue #15 field report)#22
mlstr0m merged 9 commits into
mainfrom
agent/funny-dirac-vjofbi

Conversation

@mlstr0m

@mlstr0m mlstr0m commented Sep 28, 2026

Copy link
Copy Markdown
Owner

Ports commit c349a71 from fix/issue-15-input-char-detection onto main. That commit was pushed after #16 was merged and never landed.

What it does

  • 7492866c-…c0f8 becomes a known input candidate. It is probed before any blind vendor or SIG scanning, so detection is immediate on the Switch 2 Pro Controller is not recognised #15 controller, which has no …f9 at all and streams its input from …f8.
  • …f8 is probe-validated rather than trusted outright.
  • Characteristics that can't notify or indicate are skipped everywhere. This includes a pinned ble.input_char, or …f9 itself, when it turns out to be write-only: the bridge now falls through to probing instead of failing the session.
  • 4 new tests: known alternate adopted and probed first, non-notifiable known characteristic skipped, non-notifiable pin ignored, and the updated probe order.

Adapted while porting onto v1.3.0

  • README: keeps v1.3.0's characteristics table and limitations. …f8 is notify-only, and ndeadly lists it as the Pro Controller input. The Switch 2 Pro Controller is not recognised #15 facts are folded in, and the resolution steps now sit right under the table.
  • Code comment: no longer calls …f8 the output characteristic.

The original author is kept, with (cherry picked from commit …) in the message.

Tests: 147 bridge checks and 52 DSU checks pass, and ruff is clean.

The branch also carries a merge of its previous head, which was already squash-merged in #21; the diff against main is only this port. Once this is merged, fix/issue-15-input-char-detection can be deleted.

🤖 Generated with Claude Code

https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs

Generated by Claude Code

claude and others added 9 commits September 27, 2026 18:41
@claude
Fix C/Capture swap, read factory stick calibration, hardening (v1.3.0)
d1ef492 
Protocol (sources cited in code and README):
- Byte 4 of the input report: Capture is 0x02 and C is 0x10, not the
  reverse. Agreed by ndeadly/switch2_controller_research (hid_reports.md),
  esp-cpp/espp PR #765 (switch2_pro_report.hpp) and the hardware capture
  in issue #14. Also fixes Capture -> DSU Touch.
- Read the factory stick calibration over the command channel
  (649d4ac9-...f005 / replies c765a961-...) and light player LED 1.
  Frame format, SPI addresses 0x130A8/0x130E8 and block layout adapted
  from kennethreitz/switch2bridge-macos@975f329 (issue #14, MIT), based on
  darthcloud/BlueRetro#1249; addresses match SDL's Switch 2 driver.
  Best effort: a silent controller keeps the nominal 2048 range.
- README: 7492866c-...c0f8 is notify-only, document the command channel.

Hardening:
- mappings.json written atomically (temp file + os.replace)
- "dsu.enabled" must be a boolean (bool("false") was True)
- warn when the DSU server is bound to a non-loopback address
- DSU socket: drop SO_REUSEADDR (useless for UDP; on Linux it let a
  second server silently share the port)
- worker -> UI messages taken atomically; stick hysteresis read under lock
- asyncio.get_running_loop(); logging set up at startup, not import

Build: Python 3.10+ (bleak 3 requires it), major-version upper bounds,
py2app moved to requirements-build.txt, CI matrix 3.10/3.12/3.13,
DMG README explains the macOS 15+ Gatekeeper "Open Anyway" flow.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs
@claude
Merge branch state already squash-merged in #18
7968c11
@claude
Add a manual Release workflow and the v1.3.0 release notes
7c155e7 
Actions → Release → Run workflow (tag input): checks the tag matches
APP_VERSION and doesn't exist yet, runs the tests, builds the DMG, then
creates the tag and the GitHub release with the DMG attached, using
.github/release-notes/<tag>.md as the body with {{SHA256}} replaced by
the DMG checksum.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs
@claude
Merge remote-tracking branch 'origin/main' into agent/funny-dirac-vjofbi
edd080b
@claude
Tests: keep setup_logging out of ~/Library/Logs; drop unused set_address
8870366 
Follow-up to #17:
- the setup_logging tests now point LOG_DIR at a temp dir and detach
  their handler afterwards, instead of creating controller2.log in the
  real ~/Library/Logs/Switch2Bridge and leaving a handler on bridge.log
- Mappings.set_address() had no caller outside its own test

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs
@claude
Tidy the repo: lint in CI, cleaner .gitignore, README dev docs
5c42829 
- ruff (pyflakes, bugbear, pyupgrade) configured in ruff.toml and run as
  a CI job; fixed its four findings: `raise ... from None` in config
  validation, `socket.timeout` -> `TimeoutError` (alias since 3.10)
- .gitignore: drop duplicated entries, English comments, ignore .ruff_cache
- README: project structure lists .github/ (CI, release workflow, notes)
  and ruff.toml; new Development section (tests, lint, release steps);
  DriverKit is a system extension needing a paid developer account, not
  "kernel-level" / "not possible"

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs
@claude
Merge branch state already squash-merged in #20
1a29366
@mlstr0m @claude
Promote …f8 to a known input characteristic (issue #15 field report)
3be573e 
The reporter's controller adopted 7492866c-ec3e-4619-8258-32755ffcc0f8 and
has no …f9 at all. So …f8 is now a known candidate, probed before any blind
vendor/SIG scanning, which makes detection immediate on that revision
instead of luck-of-the-ordering.

Being less sure of its role than …f9, it is probe-validated rather than
trusted outright, and characteristics that don't advertise notify/indicate
are skipped everywhere — including a pin or a known UUID that turns out to
be write-only, which now falls through to probing instead of failing the
session.

(cherry picked from commit c349a71)

Ported onto v1.3.0: the README keeps v1.3.0's characteristics table and
limitations (…f8 is notify-only; ndeadly lists it as the Pro Controller
input) and folds in the #15 facts; the code comment no longer calls …f8
the output characteristic.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs
@claude
Merge branch state already squash-merged in #21
561a582
@mlstr0m
mlstr0m merged commit 95120b1 into main Sep 28, 2026
5 checks passed
mlstr0m pushed a commit that referenced this pull request Sep 28, 2026
@claude
Release v1.4.0: bump APP_VERSION and add release notes
9c5cc5c 
Covers #17 (--config, ble.address, per-config logs), #20, #21 (lint in
CI) and #22 (…f8 known input characteristic).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs
mlstr0m pushed a commit that referenced this pull request Sep 28, 2026
@claude
Merge branch state already squash-merged in #22
b001072
@mlstr0m mlstr0m mentioned this pull request Sep 28, 2026
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@mlstr0m @claude