A scheme or workspace with a control character in a value is written as malformed XML, and XcodeProj can't read it back

未关闭 适合新手
#1,198 0 条评论 1 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
2/5
预计耗时
1-3 小时
新手友好度
85/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
swift
领域
tooling

调研方向

从 Sources/XcodeProj/Extensions/AEXML+XcodeFormat.swift 以及 xmlXcodeFormat 中属性和字符串值的转义处理入手。issue 描述了 Tests/XcodeProjTests/Extensions/XcodeFormatForbiddenCharacterTests.swift 中的测试;用它们检查禁止的控制字符、有效的控制字符,以及 scheme/workspace 往返转换。完成标准是写出的 XML 仍然可读,且完整的 swift test 测试套件通过。

由索引模型根据 Issue 内容生成。

描述

Context 🕵️‍♀️

XCScheme.write(path:override:), XCWorkspaceData.write(path:override:) and their dataRepresentation() render
through AEXMLDocument.xmlXcodeFormat (Sources/XcodeProj/Extensions/AEXML+XcodeFormat.swift). It escapes every
attribute value with AEXML's String.xmlEscaped, which handles &, <, >, ', ", \n, \r and \t. XML 1.0
also forbids, outright, the other C0 controls (U+0001–U+0008, U+000B, U+000C, U+000E–U+001F) and U+FFFE and U+FFFF.
No escape can write them; &#27; is malformed too. xmlEscaped copies them through, so the file XcodeProj writes is
not well-formed, and XCScheme(path:) / XCWorkspaceData(path:) throw on it.

The values come from callers: an environment variable's value, a pre- or post-action's scriptText, a command-line
argument, a file reference's path. Tools that generate schemes from a manifest pass such strings straight through.
#530 is an XcodeGen pre-action script that reached this writer. A terminal colour code pasted into a value, with a
raw ESC (U+001B), is the likeliest way in.

What 🌱

import PathKit
import XcodeProj

let launch = XCScheme.LaunchAction(
    runnable: nil,
    buildConfiguration: "Debug",
    environmentVariables: [XCScheme.EnvironmentVariable(variable: "PROMPT", value: "\u{1B}[31m$ \u{1B}[0m", enabled: true)]
)
let path = Path("App.xcscheme")
try XCScheme(name: "App", lastUpgradeVersion: "2600", version: "1.7", launchAction: launch).write(path: path, override: true)
_ = try XCScheme(path: path)
// throws: Error Domain=NSXMLParserErrorDomain Code=9 "(null)"
//   UserInfo={NSXMLParserErrorLineNumber=18, NSXMLParserErrorColumn=22, NSXMLParserErrorMessage=invalid character in attribute value}

Line 18 of App.xcscheme, through cat -v: value = "^[[31m$ ^[[0m". A workspace does the same: a FileRef at
.group("a\u{1}b.swift") is written as location = "group:a^Ab.swift", and XCWorkspaceData(path:) throws code 9
at line 5. With "red $ " and "ab.swift" instead, both files read back. Every one of the 30 code points fails the
same way.

Proposal 🎉

Replace each of those characters with U+FFFD in xmlXcodeFormat, after xmlEscaped:

--- a/Sources/XcodeProj/Extensions/AEXML+XcodeFormat.swift
+++ b/Sources/XcodeProj/Extensions/AEXML+XcodeFormat.swift
@@ -124,7 +124,7 @@ extension AEXMLElement {
         func print(key: String, value: String) {
             xml += "\n"
             xml += indent(withDepth: parentsCount)
-            xml += "\(key) = \"\(value.xmlEscaped)\""
+            xml += "\(key) = \"\(value.xcodeFormatEscaped)\""
         }
 
         if !attributes.isEmpty {
@@ -156,7 +156,7 @@ extension AEXMLElement {
                 // insert string value and close element
                 xml += ">\n"
                 xml += indent(withDepth: parentsCount - 1)
-                xml += ">\n\(string.xmlEscaped)"
+                xml += ">\n\(string.xcodeFormatEscaped)"
             }
         }
 
@@ -190,3 +190,21 @@ extension AEXMLElement {
         return indent
     }
 }
+
+extension String {
+    /// `xmlEscaped`, with U+FFFD in place of each character XML 1.0 cannot hold at all, even as a reference:
+    /// a C0 control other than tab, line feed and carriage return, U+FFFE or U+FFFF. A file holding one is not
+    /// well-formed, so neither XcodeProj nor any other XML reader can load it.
+    fileprivate var xcodeFormatEscaped: String {
+        var scalars = String.UnicodeScalarView()
+        for scalar in xmlEscaped.unicodeScalars {
+            switch scalar {
+            case "\u{0}"..."\u{8}", "\u{B}", "\u{C}", "\u{E}"..."\u{1F}", "\u{FFFE}", "\u{FFFF}":
+                scalars.append("\u{FFFD}")
+            default:
+                scalars.append(scalar)
+            }
+        }
+        return String(scalars)
+    }
+}

This loses the character, but the file stays loadable. The alternative is to throw from write and
dataRepresentation, which would mean xmlXcodeFormat throwing too. Output without these characters is unchanged.

I have an XCTest file, Tests/XcodeProjTests/Extensions/XcodeFormatForbiddenCharacterTests.swift, with four tests:
the scheme and the workspace above, all 30 code points in an attribute value, and a control (&, <, >, quotes,
tab and newline, which round-trip today). On main the three tests fail (32 assertions) and the control passes.
With the patch all four pass, and so does the full swift test: 387 XCTest and 95 Swift Testing tests. Happy to
open a PR with both.

AEXML's own xml output has the same gap in String.xmlEscaped (4.7.3, unchanged on its master). I have not
reported it there.

Environment. XcodeProj main at 948280a (AEXML 4.7.3). Xcode 27.2 (Swift 6.4.2), macOS 27.2, arm64.

Found while checking whether XML written by Swift packages parses.

主要语言
Swift
星标
2.2k
派生
356
平均合并
5 小时 54 分钟
30 天内合并 PR
8

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 有 Pull Request 模板
  • 没有贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

tuist/XcodeProj 的其他 Issue

查看 tuist/XcodeProj 的全部 Issue

相似的 Issue

更多 Swift Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。