Edward Diener <eldlistmailingz@xxxxxxxxxxxxxx> writes:
> On 8/31/2011 3:07 PM, Ian Lance Taylor wrote:
>> Edward Diener<eldlistmailingz@xxxxxxxxxxxxxx> writes:
>>
>>> Attempting to compile this small example program with gcc, with
>>> SOME_MACRO defined on the command line:
>>>
>>> #if !defined(SOME_MACRO)
>>> #define MACRO_1() 1
>>> #else
>>> #define MACRO_1()
>>> #endif
>>>
>>> #if defined(SOME_MACRO)
>>> int main() { return 0; }
>>> #elif MACRO_1() == 1
>>> int main() { return 1; }
>>> #endif
>>>
>>> gives the error:
>>>
>>> eliftest.cpp:9:17: error: operator '==' has no left operand
>>>
>>> "g++" -ftemplate-depth-128 -O0 -fno-inline -Wall -pedantic -g
>>> -Wno-variadic-macros -DSOME_MACRO -c "eliftest.cpp"
>>>
>>> If SOME_MACRO is defined, it seems to me that gcc should not be
>>> expanding the expression 'MACRO_1() == 1' and reporting an error.
>>
>> gcc's behaviour matches my reading of the C99 standard. The standard
>> says that each #if/#elif condition is checked. It doesn't say that the
>> compiler should stop checking conditions after it finds one that is
>> true. Instead, it says that the first group which is true is processed.
>
> You may be correct but then clearly #if - #elif - #endif works
> differently than #if - #else - #if - #endif - #endif, and that seems
> odd. Also I would have thought that an #if branch taken would have
> precluded the evaluation of an #elif statement and branch just as it
> precludes the evaluation of an #else branch.
The standard could have been written as you describe, but by my reading
it was not.
Ian
