Closed olebole closed 1 year ago
Thank you for reporting this. Yes, I believe there are explicit endian marks in the code, so that's not portable.
Note to self: re-implement test by only testing the type (float
, int
, etc.) and size (4 bytes, 8 bytes, etc.) independently of any endianness.
On a Debian system with s90x architecture (IBM zSystems, big endian, 64 bit), we get the following test failure:
Full test log here. I think this is a wrong endianness mark somewhere.