I built the Ruby MongoDB driver on a Linux Z system (big endian) and have found two issues related to endianness.

etx/lib/native.c:
rb_float_to_bson: need to convert from big endian to little endian
rb_object_id_generator_next: modified to support big endian

spec/bson/timestamp.rb
In the "#to_bson/#from_bson" test, the pack as was specified as ("l2") instead of (BSON::Int32::PACK * 2)

I also encounter 1 issues in the benchmarking code that was generic to all processor architectures. Here the FalseClass was being incorrectly bench marked against a BSON::Binary rather than a BSON::Boolean

I have built the MongoDB driver on both x86 and z390 with the changes and all test/benchmarks now run successfully through to completion.

Could you please suggest the next steps to incorporate these changes into an upcoming patch?

Thanks