I am seeing test segfaults on the Fedora aarch64 builders with soplex 7.0.0. The reason is that the copy constructor for SoPlexBase calls operator=, which compares _rationalLP to nullptr at line 1546 of src/soplex.hpp, but the copy constructor did not initialize _rationalLP, so it contains arbitrary bytes. On the x86_64 builders, we always seem to get lucky; _rationalLP is filled with zeroes. This PR ensures that is the case on all architectures.
I am seeing test segfaults on the Fedora aarch64 builders with soplex 7.0.0. The reason is that the copy constructor for SoPlexBase calls
operator=
, which compares_rationalLP
tonullptr
at line 1546 of src/soplex.hpp, but the copy constructor did not initialize_rationalLP
, so it contains arbitrary bytes. On thex86_64
builders, we always seem to get lucky;_rationalLP
is filled with zeroes. This PR ensures that is the case on all architectures.