Compiler/TMS320F28335: Issue on automatic "Enum" definition by the compiler

Hello George,

(just for information, there is a mistake in my sentence "the compiler displays a WARNING because the test is always right.", sorry).

I already read this section in the compiler manual and I have understood that the type of all "enum" values is defined according to the first value. 

And my question was just to know if there is a compiler option to disable this automatic definition of "enum" values and if it's possible to pre-define a type for all "enum" declaration.

For example: typedef enum{ VALUE1 = 0, VALUE2, VALUE3}Values

    with VALUE1/VALUE2/VALUE3 defines as "INT" and not "UNSIGNED INT".

In this case, the WARNING coming from a test always right (VALUE2 >= 0) would disappear. This is what I want.

Do you have a solution please?

Thank you very much for your support.

Best Regards,

Alexandre.

   

George Mock said:

... please submit a test case as described in the article How to Submit a Compiler Test Case.

The following can demonstrate the warning:

#include <stdbool.h>

typedef enum
{
    VALUE0,
    VALUE1,
    VALUE2
} values_t;

values_t the_value = VALUE1;

bool validate (const values_t value)
{
    return (value >= VALUE0) && (value <= VALUE2);
}

int main(void)
{
    return validate (the_value);
}

1) If compile using ti-cgt-c2000_18.1.2.LTS with --relaxed_ansi then get a warning which suggests values_t has been given an unsigned type:

Invoking: C2000 Compiler
"/home/mr_halfword/ti/ccs810/ccsv8/tools/compiler/ti-cgt-c2000_18.1.2.LTS/bin/cl2000" -v28 -ml -mt --float_support=fpu32 --include_path="/home/mr_halfword/workspace_v8/C2000_enum_type" --include_path="/home/mr_halfword/ti/ccs810/ccsv8/tools/compiler/ti-cgt-c2000_18.1.2.LTS/include" -g --c99 --relaxed_ansi --diag_warning=225 --diag_wrap=off --display_error_number --preproc_with_compile --preproc_dependency="main.d_raw"  "../main.c"
"../main.c", line 14: warning #188-D: pointless comparison of unsigned integer with zero
Finished building: "../main.c"

2) Whereas if compile using --strict_ansi don't get the warning:

Invoking: C2000 Compiler
"/home/mr_halfword/ti/ccs810/ccsv8/tools/compiler/ti-cgt-c2000_18.1.2.LTS/bin/cl2000" -v28 -ml -mt --float_support=fpu32 --include_path="/home/mr_halfword/workspace_v8/C2000_enum_type" --include_path="/home/mr_halfword/ti/ccs810/ccsv8/tools/compiler/ti-cgt-c2000_18.1.2.LTS/include" -g --c99 --strict_ansi --diag_warning=225 --diag_wrap=off --display_error_number --preproc_with_compile --preproc_dependency="main.d_raw"  "../main.c"
Finished building: "../main.c"

My reading of section 6.4.1 Size of Enum Types from SPRU514P is that for both non strict ANSI and strict ANSI modes the (signed) "int" type should be chosen before "unsigned int", assuming the enumeration values fit into an int.

To Chester Gillon ... Thank you for that test case.

To  Alexandre Contrepois ... You have never showed us exactly which warning you see.  Thus, I remain concerned that the warning in the test case from Chester is not the same one.  I'd appreciate if you would submit the test case I requested earlier.  Or, at least confirm whether the warning in Chester's test case is a match.

Thanks and regards,

-George

To George, thank you for your example, this is exactly the same warning (I use the -relaxed_ansi).

To both of you, "the_value" is defined as an unsigned type automatically because the "enum" starts at 0 (by default).

So, the test "value >= VALUE0" is always right and the compiler displays the WARNING: "pointless comparison...".

What I want is to never have this WARNING without changing the program, but adjusting the compiler options to have something like: all "enum" are defined as "int" by default (even if it starts at 0)

Do you see what I mean?

Thank you for your support guys.

Best Regards,
Alexandre.