Open DJDavies2 opened 1 week ago
I don't know if there is a solution for this. I think the code as it stands isn't standard conforming; the implementation relies on this:
stridesf = array_strides(data)
in atlas_Field_wrap pass data by reference but there is nothing in the Fortran standard that guarantees that; it can be passed via copy in/copy out and it seems that this is what nag is doing.
Sorry, I should clarify, the test failure above is actually at atlas test failure. So perhaps this should be opened there. However I think the fckit code is problematic.
What happened?
This test fails like this:
105: /spice/scratch/frwd/cylc-run/mi-be984/work/1/git_clone_atlas/atlas/src/tests/field/fctest_field_wrap.F90:121: warning: FCTEST_CHECK_EQUAL(data(i,j,l) , real(1000i+100j+10*k+l,c_double) ) 105: --> [ 1.1210000000000000E+03 != 1.1120000000000000E+03 ]
What are the steps to reproduce the bug?
Build with NAG and run this ctest.
Version
head of develop
Platform (OS and architecture)
Linux, nagfor
Relevant log output
No response
Accompanying data
No response
Organisation
Met Office