Sign-out: reach other tabs when storage refuses the sign-out's writes even after the wipe

オープン
  • 何か問題がありますか？
#1,960 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 2 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
15/100
issue の種類
機能追加
明瞭さ
説明が足りない
活発さ
活発
技術スタック
typescript
領域
frontend

調査の方向性

Start with the API explorer section of DESIGN.md and the sign-out code that writes Studio:ExplorerAuthEpoch and Studio:AppSignOut to localStorage, as described in #1956. Before any change, decide whether a BroadcastChannel fallback is worth its partial coverage, since the issue itself calls that a design call. Done means that decision is recorded and, if approved, the live-tab path is specified and tested.

索引モデルが issue の本文から書いたものです。

説明

What happens

A sign-out reaches other tabs through localStorage. It bumps the API explorer's sign-out generation (Studio:ExplorerAuthEpoch) and writes the Studio:AppSignOut announcement. #1956 makes both survive the sign-out's own storage wipe, and retries each once after the wipe has made room. Two cases are left for a store that refuses even that retry:

  1. Generation refused twice. If both the collapse and its retry are refused, the old generation stays current. A tab that missed the live event and reloads keeps an explorer credential stamped before the sign-out. This tab writes the owed collapse on its next entity sign-out, but a reload forgets that it is owed.
  2. Announcement refused twice. It is never sent. #1956 does this on purpose: an announcement sent late could end a session the user has signed back into in another tab (see the API explorer section of DESIGN.md).

Both need a store that refuses writes right after it has been emptied, such as a broken quota, a storage policy, or a private-mode quirk. Neither happens on a normal sign-out.

  1. A sign-in marker refused twice (from #1969). If storage refuses both removing and blanking a signed-out entity's marker, #1969 retries the removal later. If another tab signed in again at the same URL in the meantime, the retry can delete that new marker, because the stored value is the same either way. When the refused value couldn't be read, the retry deletes whatever is stored. Either way the entity just looks signed out on the next load.

Possible fix

Add a second cross-tab channel for when storage refuses every write, for example a BroadcastChannel the sign-out posts to. It reaches tabs that are running, not tabs that are reloading, so it covers the live path only. A durable fix for a store that won't take writes may not exist. Whether a new channel is worth that partial coverage is a design call.

Review-Deferral: 22601929e2fc387e severity=nit category=security
Review-Deferral: 6f2a8235a818a050 severity=nit category=reliability
Review-Deferral: a09380796d04a58f severity=nit category=data-integrity

Raised by the cross-model review of #1956 (#1946).

🤖 Generated by Anthropic Claude Code (Claude Opus 5.5); posted via @dawsontoth.

主要言語
TypeScript
スター
5
フォーク
4
平均マージ
1日 19時間
マージ済み PR（30日）
144

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。

HarperFast/studio のほかの issue

HarperFast/studio の issue をすべて見る

似ている issue

TypeScript 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。