yhirose/cpp-httplib v0.48.0

on GitHub

What's Changed

Security fixes

• Complete the IP-host certificate identity fix from v0.47.0 for the Mbed TLS and wolfSSL backends. An IP-literal host is now authenticated only via a matching iPAddress SAN, never via the certificate's Common Name (RFC 9110) — matching what the OpenSSL backend already enforces through X509_check_ip. Previously these backends fell back to the CN when no IP SAN matched, and recognized IPv4 only; now IPv6 (16-byte) iPAddress SANs are matched as well, and the CN fallback is skipped for both IPv4 and IPv6 literal hosts (#2476)

Improvements

• Replace the strtod-based from_chars for double with a hand-written, locale-independent parser. The only double parsed by the library is the HTTP quality value; strtod reads the decimal separator from the global C locale, so an embedder calling setlocale(LC_ALL, "") into a comma-decimal locale would mis-parse q-values. The new parser always treats . as the decimal separator and is allocation-free (Fix #2475)
• Fix OpenSSL 4.0 deprecation warnings: fetch CA store objects via the thread-safe X509_STORE_get1_objects() (OpenSSL 3.3+) and extract the subject CN via X509_NAME_get_index_by_NID()/X509_NAME_get_entry() instead of the deprecated X509_STORE_get0_objects() and X509_NAME_get_text_by_NID(). Older OpenSSL, BoringSSL, and LibreSSL keep using the get0 path. Verified warning-free against OpenSSL 4.0.1, 3.6.2, and 3.0

Behavior changes

• decode_query_component() now uses strict hex parsing for percent-escapes, consistent with decode_uri_component() and decode_path_component(). A % followed by non-hex characters (e.g. a sign or whitespace such as %-1, %+5, % 5) is passed through literally instead of being accepted as a valid escape (#2472)

Internal

• OSS-Fuzz: add new fuzzer targets for multipart parsing (#2473)

Full Changelog: v0.47.0...v0.48.0