[dbnk0]

Summary

Four small fixes for building and running ipfixprobe on embedded or cross-compiled targets such as OpenWrt:

• An opt-out for the QUIC plugin, which is the only reason OpenSSL is needed.
• A configurable path to ur_processor.sh.
• Build fixes for 32-bit platforms.
• A fix in the TLS plugin, which allocated the wrong UniRec array.

The default build works as before.

Details

• cmake: add ENABLE_PROCESS_QUIC option
  The QUIC process plugin is the only part of ipfixprobe that uses OpenSSL (to decrypt QUIC Initial packets). Because the plugin is always built, OpenSSL is a hard build and runtime dependency of the whole project. The new ENABLE_PROCESS_QUIC option defaults to ON. OpenSSL is only looked up when the option is on. This lets ipfixprobe be built for small devices that don't ship OpenSSL.

• cmake: look up ur_processor.sh instead of hard-coding its path
  The UniRec field generator always ran from /usr/bin/nemea/ur_processor.sh on the build host. In cross builds the script is staged somewhere else, and the build host may not have Nemea installed at all. The script is now found with find_program(). /usr/bin/nemea stays as a search path, so existing setups keep working, and the location can be overridden with -DUR_PROCESSOR=<path>.

• process: fix build on 32-bit targets
  On 32-bit targets, size_t is a different type from uint64_t and unsigned long, which breaks the build in two places:

  • tls_parser.cpp: quic_get_variable_length() takes the offset as uint64_t&, but parse_quic_user_agent() passed a size_t. The offset is now declared as uint64_t.
  • tls.cpp: std::min() was called with a size_t and a 99UL literal, so template argument deduction failed. It now uses std::min<size_t>.

• tls: allocate the TLS_EXT_TYPE array before filling it
RecordExtTLS::fill_unirec() allocated F_QUIC_TLS_EXT_TYPE but then filled F_TLS_EXT_TYPE. The allocation failed silently because that field isn't in the TLS template. ur_array_set() then grew the array one element at a time, moving the rest of the record on every element. The correct field is now allocated. The exported data doesn't change; only the needless reallocations go away.

---

✅ Checks

• Relevant documentation was updated (if needed)
• CI pipeline passes (build, tests, lint, static analysis)
• New code follows clang-tidy rules
• The change is clearly related to a specific issue / requirement
• Documentation (e.g. Doxygen) is updated if needed
• The PR is rebased on the latest master
• No unrelated changes are included

---