[patch,fortran] Revamp type enumerators

[Jerry DeLisle]

On 10/17/2010 07:39 AM, Jerry DeLisle wrote:
> On 10/17/2010 07:11 AM, Daniel Franke wrote:
>> On Sunday 17 October 2010 16:03:14 Jerry DeLisle wrote:
>>>>> -{
>>>>> - GFC_DTYPE_UNKNOWN = 0,
>>>>> - GFC_DTYPE_INTEGER,
>>>>> - /* TODO: recognize logical types. */
>>>>> - GFC_DTYPE_LOGICAL,
>>>>> - GFC_DTYPE_REAL,
>>>>> - GFC_DTYPE_COMPLEX,
>>>>> - GFC_DTYPE_DERIVED,
>>>>> - GFC_DTYPE_CHARACTER
>>>>> +{ BT_UNKNOWN = 0, BT_INTEGER, BT_LOGICAL, BT_CHARACTER, BT_REAL,
>>>>> BT_COMPLEX,
>>>>> + BT_DERIVED, BT_CLASS, BT_PROCEDURE, BT_HOLLERITH, BT_VOID
>>>>>
>>>>> }
>>>>> -dtype;
>>>>> -
>>>>> +bt;
>>>>
>>>> Isn't changing the values of the enums an ABI change?
>>>
>>> I was trying to avoid that, but probably got carried away. How about I
>>> commit the change to the fortran-exp branch so that it gets merged into
>>> 4.7 when we plan to break ABI anyway?
>>
>> Wouldn't it be sufficient to re-order the BT_* to match the GFC_DTYPE_*?
>> Only BT_CHARACTER is out of order, BT_CLASS and on only add more identfiers.
>> So, shouldn't this definition work, as in not breaking anything?
>>
>> + { BT_UNKNOWN = 0, BT_INTEGER, BT_LOGICAL, BT_REAL, BT_COMPLEX,
>> + BT_DERIVED, BT_CHARACTER, BT_CLASS, BT_PROCEDURE, BT_HOLLERITH, BT_VOID }
>>
>
> Possibly, but I would want to test. One way to test would be to copy the before
> patch libraries into the after patch build directory and run the testsuite.
> Regardless, this is a fairly intrusive patch even though mostly mechanical.
> Anyone else have any opinions?
>

Daniel's suggestion appears to work.  After some testing I idetified a minor 
unrelated issue with backward compatibility of an empty STOP statement.  It is 
unrelated to the BT enumerator consolidation patch. (I will submit a patch for 
the STOP issue separately)

With Daniel's change above, OK for trunk?

Regards,

Jerry