[marty1885]

The current multi part parser does not handle short invalid multipart bodies correctly. For the multipart body A\r\nA (which is invalid according to RFC). The parser will attempt to read at index pos2 - 4 which becomes -1 therefor a OOB read.

Also a reordering in RequestImpl to fix a harmless bug that pos < input.length() is evaluated after the read input[pos] == '?'. Which is fine in reality because the tailing \0 in C strings. But still wroth fixing.

[Copilot]

🟡 Changes recommended

The regression test can pass despite the out-of-bounds read, so it needs AddressSanitizer coverage or another deterministic check.

1 open issue

• Add deterministic coverage for the parser's out-of-bounds read · New

What changed in this PR

This PR hardens request parsing against out-of-bounds reads in malformed multipart bodies and parameter input.

Changes:

• Reject empty multipart boundaries and insufficient data before boundary-related reads.
• Check parameter input bounds before accessing characters.
• Add regression cases for nearby boundaries and a boundary at the end of the body.

File	Description
lib/​src/​MultiPart.cc	Adds bounds checks around multipart boundary processing.
lib/​src/​HttpRequestImpl.cc	Moves bounds checks before character access in parameter parsing.
lib/​tests/​unittests/​MultiPartParserTest.cc	Adds malformed multipart rejection tests.

_{Review effort: Balanced}