[200_60] Goldfish Scheme 支持忽略 shebang

[da-liii]

功能描述

在 load_file_1 函数中添加 shebang 行跳过逻辑，使 Goldfish Scheme 能够正确执行以 #! 开头的脚本文件。

实现细节

• 在加载文件内容后、解析执行前，检查文件内容是否以 #! 开头
• 如果是，则跳过第一行（shebang 行）
• 处理有换行符和无换行符的情况

使用示例

用户现在可以创建如下脚本并直接执行：

#!/usr/bin/env goldfish
(display "Hello, World!")

相关文档

• devel/200_60.md - 任务详细说明

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR adds shebang (#!) support to Goldfish Scheme by modifying load_file_1 in src/s7.c, allowing scripts that start with #!/usr/bin/env goldfish to be executed directly on Unix-like systems.

• The implementation checks whether the loaded port is a string_port with data beginning with #!, then advances the read position past the first line before parsing begins.
• The logic is correct and safe on non-Windows platforms for normal files: read_file always creates a string_port when loading via load_file_1 (which passes max_size = -1).
• Windows gap: On Windows, read_file unconditionally creates a file_port (the #else branch sets port_data = NULL). The is_string_port guard prevents a crash, but the shebang line is never skipped, so shebang scripts will fail to parse on Windows without any error message explaining why.
• No automated test cases were added for the shebang feature. Per the project's TDD guidelines in CLAUDE.md, tests should accompany new functionality.

Confidence Score: 3/5

Safe to merge for Unix-like platforms; Windows shebang support is silently broken and no tests were added.

The core logic is correct and non-crashing on all platforms (the is_string_port guard prevents NULL dereference on Windows/pseudo-files). However, the feature silently does nothing on Windows, and there are no test cases to verify correct behavior or catch regressions — both gaps that should be addressed before or shortly after merging.

src/s7.c — specifically the Windows file_port path that is unhandled by the new shebang logic.

Important Files Changed

Filename	Overview
src/s7.c	Adds shebang-skip logic inside load_file_1; correct for the common (non-Windows) string-port path, but silently ineffective on Windows where the port is always a file_port.
devel/200_60.md	New task documentation file; accurately describes the What/Why/How of the shebang implementation. No code issues.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[load_file_1 called] --> B[fopen file]
    B --> C{fp valid?}
    C -- No --> Z[return NULL]
    C -- Yes --> D["read_file(sc, fp, name, -1, 'load')"]
    D --> E{Platform?}
    E -- "Windows (#else)" --> F["port_type = file_port\nport_data = NULL"]
    E -- "Non-Windows + size > 0" --> G["port_type = string_port\nport_data = content buffer"]
    E -- "Non-Windows + size == 0\n(pseudo-files)" --> F
    F --> J["is_string_port? → false\nShebang check SKIPPED"]
    G --> H["is_string_port? → true\nCheck data[0]=='#' && data[1]=='!'"]
    H -- No shebang --> K[port_position stays at 0]
    H -- Shebang found --> I{newline found?}
    I -- Yes --> L["port_position = pos + 1\n(skip shebang line)"]
    I -- No --> M["port_position = port_data_size\n(empty remainder)"]
    J --> N[push_input_port & parse]
    K --> N
    L --> N
    M --> N

Loading

_{Reviews (1): Last reviewed commit: "[200_60] Goldfish Scheme 支持 shebang" | Re-trigger Greptile}

[greptile-apps]

Shebang support silently skipped on Windows

The shebang-skip logic is guarded by is_string_port(port), but on Windows, read_file always creates a file_port (not a string_port) — see the #else branch starting at line 27577 in src/s7.c. This means the #! line is never skipped on Windows, and the interpreter will encounter a syntax error when loading a shebang script on that platform.

The Windows-specific branch sets port_data = NULL and port_data_size = 0, so the is_string_port guard correctly prevents a NULL-dereference crash, but the feature silently fails to work.

If Windows shebang support is needed, a separate code path for file_port (using getc/ungetc or similar) would be required. If Windows is intentionally out of scope, a comment here would clarify the limitation.

[greptile-apps]

Shebang skipped for zero-size / pseudo-files (edge case)

When fseek/ftell reports size == 0 (e.g., files under /proc), read_file also creates a file_port, not a string_port. Those files won't receive shebang handling. This is unlikely to matter in practice since /proc pseudo-files don't use shebangs, but it is a subtle inconsistency worth noting alongside the Windows case above.