Blocking driver cannot return to continuous mode; README documents a workaround that cannot be written

オープン
  • 何か問題がありますか？
#81 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

まだ誰も着手していません。

評価

難易度
3/5
見積もり時間
1〜2日
初心者へのやさしさ
74/100
issue の種類
機能追加
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
rust
領域
documentation, embedded-iot, testing

調査の方向性

Start with the blocking Tmp108 implementation at src/lib.rs:1847 and compare its mode handling with AsyncTmp108::continuous at src/lib.rs:3016-3033. Review the blocking tests, the async doctest, and the README.md and AGENTS.md references named in the issue. Done means the blocking API, cleanup and error cases, crate-level documentation, and misleading workaround text are covered by tests and docs.

索引モデルが issue の本文から書いたものです。

説明

audit:SBOS663A bug enhancement impact:documentation

The blocking driver can leave continuous conversion but cannot return to
it. README.md:185-187 says otherwise.

The gap

set_m(Mode::Continuous) appears exactly once in the crate:

src/lib.rs:3022   .modify_async(|r| r.set_m(Mode::Continuous))

That line is inside impl<I2C: AsyncI2c> AsyncTmp108<I2C> (block opens at
src/lib.rs:2474). The blocking block, impl<I2C: I2c> Tmp108<I2C> at
src/lib.rs:1847, writes only:

src/lib.rs:2224   .modify(|r| r.set_m(Mode::Shutdown))     // one_shot preparation
src/lib.rs:2234   .modify(|r| r.set_m(Mode::OneShot))      // one_shot trigger
src/lib.rs:2278   .modify(|r| r.set_m(Mode::Shutdown))     // shutdown

Config carries thermostat_mode, alert_polarity, conversion_rate
and hysteresis — no mode field — and apply_config preserves M. So
a default-features user who calls shutdown(), or one_shot() (which
self-clears to shutdown per SBOS663A §7.4.2), has no way back to
continuous conversion through this crate.

The README is wrong about the workaround

README.md:185-187:

  • AsyncTmp108::continuous is async-only. For blocking continuous-mode
    use, call Tmp108::configure(...) to set Mode::Continuous manually,

configure takes a Config, and Config has no mode field. This
workaround cannot be written. AGENTS.md:28 and AGENTS.md:112 repeat
the "no blocking equivalent" framing without proposing the impossible
remedy, so the README line is the only actively misleading one.

Datasheet context

Continuous conversion is the power-on default — the configuration reset
value is 0x1022, M = 0b10 (SBOS663A Table 8,
sources/datasheet.txt:883-889). The only documented way to restore
power-up register values without software setting M is the general call
reset command:

the TMP108 internal registers are reset to power-up values. The TMP108 does not support the general address

SBOS663A §7.3.7, sources/datasheet.txt:717

That is a broadcast to address 0x00, which resets every general-call
capable device on the bus, so it is arguably not a per-device driver's to
offer. Which leaves the blocking driver with no in-crate escape at all.

Proposed fix

Mirror the async method. The async body is four statements
(src/lib.rs:3016-3033); the blocking one is the same without .await,
and is simpler, because FnOnce avoids the AsyncFnOnce closure-syntax
trap documented in AGENTS.md gotcha #4:

pub fn continuous<F>(&mut self, f: F) -> Result<(), I2C::Error>
where
    F: FnOnce(&mut Self) -> Result<(), I2C::Error>,
{
    self.inner
        .configuration()
        .modify(|r| r.set_m(Mode::Continuous))?;

    let user_result = f(self);
    let cleanup_result = self.shutdown();
    user_result.and(cleanup_result)
}

Purely additive, and ungated — unlike the async one it needs no feature.

Cost
Item Size
Production code ~8 lines in the Tmp108 impl block
Doctest required by AGENTS.md; mirror src/lib.rs:2993-3015 minus the block_on wrapper
Rustdoc needs the # Interrupt-mode acknowledgement section — it is a read-modify-write — plus the qualified cleanup wording #77 just landed on the async one
Crate-level table add a Tmp108::continuous row at src/lib.rs:51-60; today only AsyncTmp108::continuous is listed
Unit tests mirror the async entry-failure / cleanup-error / dropped-error cases in mod tests::blocking
README.md rewrite the gotcha at 185-187; update the feature table at 108
AGENTS.md update lines 28, 101, 112 and gotcha #4's framing
tests/reexports.rs optional use-site pin
Semver additive → minor bump; feat: subject, release-plz handles it
Examples optional: examples/continuous.rs could gain a blocking arm like oneshot.rs and sensor_trait.rs already have. If it does, the CI example matrix and scripts/check-readme-snippets.sh need the extra combination.

The snippet script is unaffected as long as examples/continuous.rs
stays async-only.

Alternative considered

An additive start_continuous() that just writes M without the scoped
cleanup. Rejected: it reintroduces the "who turns it off" problem that
continuous()'s closure shape exists to solve, and the blocking port is
cheap enough that there is no reason to offer the weaker API.

Provenance

Raised by the datasheet compliance audit of PR #77. Two audit agents
reached it from opposite ends — one looking for documented recovery paths,
one for facts absent from the types.

主要言語
Rust
スター
2
フォーク
7
30日以内にマージされた PR はありません

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

OpenDevicePartnership/tmp108 のほかの issue

OpenDevicePartnership/tmp108 の issue をすべて見る

似ている issue

Rust の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。