From 2869e510c185220268fb9d1b8ade2e8dfd9d5f99 Mon Sep 17 00:00:00 2001
From: wujunchen <wujunchen@baidu.com>
Date: Sun, 21 Jun 2026 19:44:41 +0800
Subject: [PATCH] fix(opencode): strip all leading BOMs in Bom.split (#33092)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Bom.split() removed only the first U+FEFF via text.slice(1), so content
with multiple consecutive leading BOMs kept the extras. Because join()
re-runs split() before re-adding a single BOM, each write/edit cycle
dropped at most one extra BOM, so a file with N leading BOMs needed
several edit passes to normalize to one.

Align split() with the existing splitBom() in
packages/core/src/file-mutation.ts and strip all leading BOMs via
/^+/. join() is unchanged; multi-BOM content now normalizes to
exactly one (or zero) BOM in a single pass, fixing the write/edit/
apply_patch callers transitively.

Add unit tests for split()/join() (the util had none) covering no-BOM,
single, multiple consecutive, and non-leading BOMs.

Fixes #33092

Claude-Session: https://claude.ai/code/session_01444Qm3LDDrcHSy1HtETM6G
---
 packages/opencode/src/util/bom.ts       |  4 +--
 packages/opencode/test/util/bom.test.ts | 44 +++++++++++++++++++++++++
 2 files changed, 46 insertions(+), 2 deletions(-)
 create mode 100644 packages/opencode/test/util/bom.test.ts

diff --git a/packages/opencode/src/util/bom.ts b/packages/opencode/src/util/bom.ts
index f015651e97fd..16b578eeccaf 100644
--- a/packages/opencode/src/util/bom.ts
+++ b/packages/opencode/src/util/bom.ts
@@ -5,8 +5,8 @@ const BOM_CODE = 0xfeff
 const BOM = String.fromCharCode(BOM_CODE)
 
 export function split(text: string) {
-  if (text.charCodeAt(0) !== BOM_CODE) return { bom: false, text }
-  return { bom: true, text: text.slice(1) }
+  const stripped = text.replace(/^\uFEFF+/, "")
+  return { bom: stripped.length !== text.length, text: stripped }
 }
 
 export function join(text: string, bom: boolean) {
diff --git a/packages/opencode/test/util/bom.test.ts b/packages/opencode/test/util/bom.test.ts
new file mode 100644
index 000000000000..7fffb6beb19f
--- /dev/null
+++ b/packages/opencode/test/util/bom.test.ts
@@ -0,0 +1,44 @@
+import { describe, expect, test } from "bun:test"
+import * as Bom from "../../src/util/bom"
+
+const BOM = "\uFEFF"
+
+describe("Bom.split", () => {
+  test("returns text unchanged when there is no BOM", () => {
+    expect(Bom.split("hello")).toEqual({ bom: false, text: "hello" })
+  })
+
+  test("strips a single leading BOM", () => {
+    expect(Bom.split(BOM + "hello")).toEqual({ bom: true, text: "hello" })
+  })
+
+  test("strips all consecutive leading BOMs", () => {
+    expect(Bom.split(BOM.repeat(4) + "hello")).toEqual({ bom: true, text: "hello" })
+  })
+
+  test("does not strip a BOM that is not at the start", () => {
+    expect(Bom.split("hello" + BOM)).toEqual({ bom: false, text: "hello" + BOM })
+  })
+
+  test("handles an empty string", () => {
+    expect(Bom.split("")).toEqual({ bom: false, text: "" })
+  })
+})
+
+describe("Bom.join", () => {
+  test("adds a BOM when requested", () => {
+    expect(Bom.join("hello", true)).toBe(BOM + "hello")
+  })
+
+  test("omits the BOM when not requested", () => {
+    expect(Bom.join("hello", false)).toBe("hello")
+  })
+
+  test("normalizes multiple existing leading BOMs to exactly one", () => {
+    expect(Bom.join(BOM.repeat(4) + "hello", true)).toBe(BOM + "hello")
+  })
+
+  test("strips all existing leading BOMs when none is requested", () => {
+    expect(Bom.join(BOM.repeat(3) + "hello", false)).toBe("hello")
+  })
+})