Decode piped stdin as UTF-8 and strip any BOM (PowerShell 5.1 compat)

[Jovenkemp]

Problem

On Windows PowerShell 5.1, piped text often arrives with a UTF-8 BOM (EF BB BF): the stock [Text.Encoding]::UTF8 $OutputEncoding emits its preamble when piping to native commands, and files written by PS 5.1 (Out-File -Encoding utf8) carry BOMs that survive re-piping. sys.stdin.read() keeps that BOM in the decoded source — U+FEFF, or  when decoded via the locale code page — so exec() fails on line 1 of every piped script:

SyntaxError: invalid non-printable character U+FEFF (<string>, line 1)

Since piping code into browser-harness is the documented (and only) usage mode, this breaks the harness entirely for PowerShell 5.1 callers.

Fix

Read stdin as raw bytes and decode with utf-8-sig, which strips one leading UTF-8 BOM when present and is byte-for-byte identical to utf-8 otherwise.

Compatibility notes (each verified against CPython)

• bash / cmd / heredoc input is unchanged. With no BOM present, utf-8-sig decodes exactly like utf-8.
• Newline translation is not lost. Text-mode stdin used to translate \r\n to \n; compile()/exec() normalize newlines themselves, including inside string literals (exec('s="""a\r\nb"""') yields 'a\nb'), so CRLF input behaves identically.
• UTF-16 stdin: deliberately out of scope. CPython rejects UTF-16 source files (SyntaxError: Non-UTF-8 code starting with '\xff' ... see PEP 263), so the harness keeps the same contract for piped source. A UTF-16 pipe now fails loudly at decode time (UnicodeDecodeError, or ValueError: source code string cannot contain null bytes) instead of producing mojibake.
• Side benefit: non-ASCII piped source now decodes as UTF-8 on Windows instead of the locale code page (cp1252), matching how Python defines source encoding (PEP 3120).

This mirrors the existing UTF-8 normalization the file already does for stdout (run.py lines 3–8).

🤖 Generated with Claude Code

---

Summary by cubic

Fix stdin decoding so piped scripts work on Windows PowerShell 5.1. We now read bytes and decode as UTF‑8 while stripping a BOM, preventing the first-line SyntaxError when piping into browser-harness.

• Bug Fixes
  • Read from sys.stdin.buffer and decode with utf-8-sig to drop a leading BOM from PowerShell 5.1 pipes.
  • No change for non-BOM input; newline handling unchanged. UTF‑16 stdin remains unsupported and now fails early.

^{Written for commit 29655f1. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 1 file

_{Re-trigger cubic}