MATLAB Matrix API/MAT-File API seems to disable extended floating-point precision for the entire process it is loaded by

My code base has C, C++ and Ada interop, and use the MATLAB Matrix API/MAT-File API to read and parse MAT files. Parts of the Ada code is reliant on extended floating point precision for e.g. ceil and floor functions to work properly. This all fell apart after loading the MATLAB shared objects. The ceil and floor functions suddenly returned weird values, and parts of the code started emitting constraint error exceptions. Is there a way around this, or do i have to separate the code utilizing the MATLAB APIs from the code that is reliant on extended floating-point precision?

Answer 1

[Jon Skjelsbæk]

Found the solution: https://github.com/ami-iit/matio-cpp. Works like a charm.