Open dschwoerer opened 4 months ago
Yeah - in the first case the test fails because of a limitation in floating point precision and isn't really a fail. Basically the numbers beyond a certain number of decimal places will be random
In the second and third case, I'm not sure but I expect that these are compiler implementation differences - the test data is generated from the Intel ifort
compiler, whereas the FortranFormat
code is based of the source from gfortran
.
The intention of the library was to emulate different compilers using configuration flags but really I think this is a low priority for most. I'm happy for this issue to remain as a bug until those flags are implemented.
The failing tests: test-failures.log.gz
full log: builder-live.log.gz
If I understood you correctly, this is a known issue. In that case feel free to close this as "won't fix".
Just looking at the first 3 failures:
I do not think that is something that should be 'fixed' ...