Bug in mpas_io_streams.F: real3dField is queried for a real4dField #1365
Conversation
… is queried. This would crash probably, but until now this section of the code is not executed.
| @@ -3074,7 +3074,7 @@ subroutine MPAS_readStream(stream, frame, ierr) | |||
|
|
|||
| else | |||
|
|
|||
| if (field_cursor % real3dField % isVarArray) then | |||
| if (field_cursor % real4dField % isVarArray) then | |||
| call mpas_dmpar_bcast_reals( mpas_io_handle_dminfo(stream % fileHandle), size(real3d_temp), real3d_temp(:,1,1)) | |||
| field_4dreal_ptr => field_cursor % real4dField | |||
| do while (associated(field_4dreal_ptr)) | |||
There was a problem hiding this comment.
@mgduda Not knowing this code, I just wanted to check that a 3D temp field indexed to first index j in the 4d field is what is intended on the next line. A few more lines down, there is a 4D temporary array, which is what I would have expected. Could you review that line?
There was a problem hiding this comment.
@mark-petersen Hey Mark, thanks for looking at this. I believe real3d_temp is correct in line 3078, since we are testing real4dField being a varArray or not. If you scroll up or down a few lines, you will find that the remainder of the code is using/deallocating real3d_temp inside "if field_cursor % real4dField % isVarArray)".
Also, if you look at line 2990 (similar piece of code for field_cursor % real3dField % isVarArray, you will see that it uses real2d_temp inside those if blocks.
There was a problem hiding this comment.
Great, thanks. That makes sense.
In mpas_io_streams.F, field_cursor % real3dField % isVarArray is queried in one place for a 4d real field. The correct behaviour would be to query field_cursor % real4dField % isVarArray.
Apparently, we do not meet the section of the code at present, hence this error never came up. The code should probably crash if we did.