audio_play and audio_stream_push accept a wrong-typed sample element when no audio device is open

オープン
#2,447 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
4/5
見積もり時間
3〜5日
初心者へのやさしさ
43/100
issue の種類
バグ
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
c
領域
backend

調査の方向性

Start with the audio builtins and their neighboring checks, then inspect tools/strict_differential.sh and the docs/BUILTINS.md rows named in the issue. Reproduce the wrong-typed sample behavior with and without an audio device, and measure the stated headless cost for a per-frame audio consumer. Done means both builtins behave as specified, the differential probe runs without opening a device and is examined on macOS, and the documentation matches a run.

索引モデルが issue の本文から書いたものです。

説明

area:gfx found-by:ci kind:silent-wrong

On phase2-numbers (and on the branch of #2442, same source here), two audio builtins look at the sample list's ELEMENTS only after they have checked that a device is open. On a machine with no audio device, which is every CI runner without SDL and every headless run, a wrong-typed sample element is accepted in every mode.

$ printf 'print of (audio_play of [0.1, "x", 0.2])\n' > a1.eigs
$ SDL_AUDIODRIVER=nosuchdriver eigenscript a1.eigs                  # 0
$ EIGS_STRICT=1 SDL_AUDIODRIVER=nosuchdriver eigenscript a1.eigs    # 0
$ printf 'print of (audio_stream_push of [0.1, "x", 0.2])\n' > a3.eigs
$ EIGS_STRICT=1 SDL_AUDIODRIVER=nosuchdriver eigenscript a3.eigs    # false

With a device open (SDL_AUDIODRIVER=dummy, after audio_open of [44100, 1]) the first one raises audio samples: element 1 is str, expected a number, as it should.

The neighbours already do it the other way round. The whole-argument check of both builtins was moved above the device check in #1007 (audio_play of 42 raises with no device), and audio_play_loop checks its elements above the device check: audio_play_loop of [[0.1, "x", 0.2], 1] raises with no device. So the same mistake is loud in one builtin and silent in the two beside it, and which one a caller gets depends on the machine.

Found through a probe of the strict differential (tools/strict_differential.sh) that needs an open device to reach this guard: it cannot run on the macOS runners. #2442 lists that probe as not examined there; this issue is the fault behind it.

Done when

  • audio_play and audio_stream_push refuse a wrong-typed sample element whether or not a device is open, with the message they give today when one is open. EIGS_STRICT=0 keeps today's quiet answer.
  • The cost on a machine with no device is stated: a check of the element types without converting the samples, or the conversion with its measured cost on a headless run of a consumer that plays audio every frame.
  • The differential's probe for this guard no longer opens a device, and is examined on the macOS runners.
  • docs/BUILTINS.md rows for the two builtins say what a wrong-typed element does with and without a device, and the sentence is true by a run.

— Claude DevBox

主要言語
C
スター
3
フォーク
7
平均マージ
3時間 28分
マージ済み PR(30日)
270

環境構築

Codespaces で開く

このプロジェクトの開発コンテナを、あなたの GitHub アカウントでブラウザ上に起動します。

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。

InauguralSystems/EigenScript のほかの issue

InauguralSystems/EigenScript の issue をすべて見る

似ている issue

C 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。