Fix POST/PUT request bodies being lost server-side

[sclaiborne]

Problem

LLHttpParse() only sets contentPresent / partialContent in the HTTP response parsing branch. The request parsing branch never sets them, so LLHttpServer''s dispatch (if(client->parser.contentPresent)) always copied an empty body into LLHttpResponse handlers. POST/PUT bodies were silently lost server-side: requestLength reported the Content-Length, but pRequest received only '\0'. Reproduced against ARsim with LLHttpServer + LLHttpResponse.

On top of that, the partial-content accumulation path in HttpServer.c was broken for multi-segment requests: after HttpShiftReceivePointer(), the parser was given the receive buffer base pointer but only the latest segment''s ReceivedDataLength, so a request whose headers and body arrive in separate TCP segments (which is what .NET HttpClient sends - headers and body in separate writes) never parsed completely. The appended null terminator was also written one byte past the received data.

Fix

• HttpParse.c: mirror the response-branch logic at the end of the request branch - if header.contentLength != 0, set contentPresent = 1 and partialContent = (returnValue + contentLength > dataLength).
• HttpServer.c:
  • Parse over the full accumulated buffer: recvLength = shift offset + newly received length (matching the client''s accumulation in HttpClient.c), and shift the receive pointer by the new segment length only.
  • Append the null terminator at data[recvLength] instead of data[recvLength+1], and reserve one byte of the receive buffer for it (as the client does).
  • Zero-initialize the TMP_alloc''d client array in LLHttpServerInit() - TMP_alloc memory is not initialized, so TCP stream state/FUB statuses started as garbage.

Tests

Added under test/:

• Parser: request with body sets contentPresent/content, request with partial body sets partialContent.
• Server: a POSTed JSON body split across two TCP segments (headers first, body second) is accumulated, parsed, and dispatched to a registered handler with the full body. Verified the new tests fail (empty body, exactly the reported symptom) with the parse fix reverted.

Test-infra repairs (needed to build/run the suite at all)

• CMakeLists.txt: the src/Ar restructure broke the CMake build - add_subdirectory(example) pointed at a directory with no CMakeLists, and the library target had no include dirs. Fixed paths and stale ../LLHttpH.h-style includes in tests.cpp/example.cpp.
• Bumped Catch2 v3.0.0-preview3 -> v3.4.0 (preview3 does not compile with current GCC) and renamed Contains -> ContainsSubstring per Catch2 v3 API.
• Added brsstrlen fallback macro for non-AR builds (used by the chunked-encoding path).
• Create the report/ directory at configure time so ctest passes on a clean checkout.

Verified per TestJenkinsfile: CMake + MinGW (32-bit, matching UDINT pointer width) build, full Catch2 suite green via ctest - 7/7 test cases, 139 assertions.

🤖 Generated with Claude Code