[Aryan9891]

Problem

As noted in Note 2 of #2299, RFC 8259 Section 7 requires characters U+0000 through U+001F inside JSON strings to be escaped. In src/json_reporter.cc, StrEscape explicitly escapes \b \f \n \r \t \" \\ but allows other C0 control characters (e.g. U+0001 or ANSI escape characters like U+001B in benchmark names or labels) to pass through unescaped, causing strict JSON parsers to reject the output.

Solution

In StrEscape, escape any remaining byte in the C0 control range (< 0x20) as \u00XX via std::snprintf. Printable characters and UTF-8 multibyte sequences remain unaffected.

Fixes #2299.

[dmah42]

can we slightly simplify this with auto uc = static_cast<unsigned char>(c); which can then be reused in the snprintf below?

[dxnnystyles]

this looks like a good fix! it'd be worth adding a regression test with something like \x00 and \x1f in a benchmark name, then checking that the generated report is valid JSON and the characters survive decoding. that would protect the behavior this PR is fixing.

[Aryan9891]

Thanks for the review and suggestions! I have updated the PR to address your feedback:

• Simplified the C0 control character handler using \�uto uc = static_cast(c);\ and reused it directly in \std::snprintf.
• Added unit tests in \ est/reporter_list_gtest.cc\ verifying that C0 control characters (\x01, \x1f) in benchmark names are escaped as \u0001\ and \u001f\ in the JSON output per RFC 8259.
• Rebased cleanly onto the latest \main\ branch.

[dxnnystyles]

thanke! I checked the revised escaping logic + regression test. the new assertions cover the expected \u0001 and \u001f output, and the unsigned-char handling looks good. that addresses my original concern 👍

[dmah42]

Thanks for the review and suggestions! I have updated the PR to address your feedback:

• Simplified the C0 control character handler using \�uto uc = static_cast(c);\ and reused it directly in \std::snprintf.
• Added unit tests in \ est/reporter_list_gtest.cc\ verifying that C0 control characters (\x01, \x1f) in benchmark names are escaped as \u0001\ and \u001f\ in the JSON output per RFC 8259.
• Rebased cleanly onto the latest \main\ branch.

whatever agent you're using (which you haven't declared per AGENTS.md) is having a hard time with escaping characters. might want to double check that.