Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
28454 Discussions

Missing error when using a non-TARGET variable in C_LOC

ereisch
New Contributor II
459 Views

When using C_LOC from ISO_C_BINDING, the standard states that items passed to C_LOC() must have the TARGET attribute.  However, when I pass in a simple BYTE variable from a COMMON structure, the compiler does not produce any warnings or errors as would be expected.

 BYTE START_ADR, END_ADR
 STRUCTURE    /TEST/
    COMPLEX*8 SOME_VAL1
    REAL*4 SOME_VAL2
 END STRUCTURE

 RECORD /TEST/ FOO

 COMMON/CTEST/
& START_ADR(8),
& FOO(10000),
& END_ADR(4)

If I create a line with C_LOC( START_ADR ), ifort treats it as valid, but gfortran reports it as a violation of the standard.  Intel's documentation on C_LOC seems to suggest it should be illegal as well unless I provide "TARGET START_ADR, END_ADR" somewhere after the variable definition.

Additionally (unrelated topic), in my implementation, I also define VOLATILE CTEST, but the standard suggests VOLATILE can only be defined on variables, not common definitions.

0 Kudos
2 Replies
andrew_4619
Honored Contributor II
440 Views
This already has a thread from 28/11/2020
0 Kudos
andrew_4619
Honored Contributor II
418 Views

https://community.intel.com/t5/Intel-Fortran-Compiler/Stand-C-LOC-and-Target-Attribute/m-p/1232351#M152898

The above was the topic link. A support ticket was raised. I am not sure about the VOLATILE which is a different topic as you note. Given your code has non-standard constructs I would think you are not using standards checking.  Intel could thus accept non-standard usage as an language extension. You would need to make an otherwise conforming test case I think. 

0 Kudos
Reply