Hacktoberfest 2026：メンテナが10月に向けて印を付けた、オープンで初心者向けの issue。 Hacktoberfest の issue を見る

Opcode facts in one X-macro table; retire the three opcode-comment gates (#1275)

オープン
  • 何か問題がありますか？
#1,320 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
5/5
見積もり時間
1週間以上
初心者へのやさしさ
28/100
issue の種類
リファクタリング
明瞭さ
おおむね明確
活発さ
活発
技術スタック
c, shell
領域
compilers, testing-qa

調査の方向性

Start by reading src/vm.h, src/chunk.c, and the three gate scripts: tools/obs_marker_check.sh, tools/obs_reader_sync_check.sh, and tools/vm_operand_width_check.sh. Trace the opcode enum, observer lookup, and verifier role table before assessing the X-macro design. Done means the generated facts replace the duplicated checks, bytecode remains identical, and the full suite, ASan, JIT checks, and calibration plant pass.

索引モデルが issue の本文から書いたものです。

説明

area:gates area:runtime-vm kind:gate-defect

Split out of #1275 (item 7, "the rest"), as its definition of done requires for rows that need a product-code change.

Problem

Three gate scripts check hand-written opcode comments against the C (1,277 lines together at 94c4c5b). None has a recorded catch of a product bug:

  • tools/obs_marker_check.sh (550): every opcode in src/vm.h has an /*obs:READS|NONE|…*/ marker (#972).
  • tools/obs_reader_sync_check.sh (448): those markers agree with the case OP_…: arms of opcode_is_observer_reader() in src/chunk.c (#915/#972).
  • tools/vm_operand_width_check.sh (279): the [kind:N] operand-width comments in src/vm.h agree with the VM's decoder sites and chunk.c's VR_* verifier table (#958).

Each property lives in two or three homes: a comment, a switch and a table. The scripts only exist to keep those homes in step.

Proposal (structural, principle 5)

One X-macro opcode table, for example src/opcodes.def, with one row per opcode carrying NAME, observer-class, operand roles/widths. The rest is generated from it by the preprocessor:

  • the enum in vm.h;
  • opcode_is_observer_reader() (a lookup of the observer-class column);
  • the verifier's role table.

The per-opcode facts then have one home. The compiler enforces the agreement that the three scripts check today, and they are deleted. What remains is at most a small check that the decoder's read_uN widths match the table's widths (the one fact the preprocessor cannot tie). Measure whether it is needed before keeping it.

Touches the opcode ABI (see the eigenscript-extend-vm skill and the operand-width-is-ABI note): the bytecode and the JIT must be byte-identical before and after.

Done when

  • Opcode name, observer class and operand roles/widths are declared in exactly one table; enum, opcode_is_observer_reader() and the verifier roles are generated from it.
  • tools/obs_marker_check.sh, tools/obs_reader_sync_check.sh and tools/vm_operand_width_check.sh are deleted along with their suite sections and tools/selftests.txt rows, or one survivor is ≤ 150 lines and the PR shows why it is needed.
  • The compiled bytecode is byte-identical before and after on the corpus: --dump, or a hash of every compiled tests/*.eigs chunk, compared old against new.
  • Full suite, ASan-with-leaks and tools/jit_diff.sh are green; make jit-smoke passes.
  • One calibration plant: flipping one opcode's observer class in the table changes opcode_is_observer_reader() and makes the suite's observer section fail. The result is recorded in the PR.
主要言語
C
スター
3
フォーク
7
平均マージ
4時間 7分
マージ済み PR（30日）
112

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。