[manoj-1407]

Summary

tomlkit accepts a few numbers that TOML does not allow:

parse("v = 1.٠")       # ARABIC-INDIC DIGIT ZERO
parse("v = 1\x0b")     # a vertical tab after the number

After checking the shape of the number, the parser passes it to int() and float(). Both accept non-ASCII digits and surrounding whitespace, so these were parsed as valid values. The parser now rejects a number that has a non-ASCII character or a whitespace character before that conversion.

The current toml-test suite has 8 invalid cases for this (invalid/float/arabic-zero-01/03/04, invalid/integer/arabic-zero-01/02, invalid/control/linetab-number-01/02/03), which tomlkit accepted. With this change all of them are rejected, and the rest of the suite still passes. I added a parametrized test to tests/test_parser.py and a changelog entry.

Agent Drafting Metadata

• Agent: Claude Code
• Model: Claude Sonnet 5.5
• Notes: Claude found this by running the toml-test cases against tomlkit and wrote the patch. I read the change, ran the full test suite, and checked the cases myself.