Eric Blake <[EMAIL PROTECTED]> writes:
>>> "../../lib/regexec.c", line 1412: warning: non-constant initializer: op "--"
>>
>> I don't understand this one. Here's the line in question:
>>
>> Idx num = --fs->num;
>>
>> and I assume Idx is size_t, which is a 32-bit unsigned integer, so this
>> code should be perfectly fine C89. Am I missing something? It appears
>> to me to be a compiler bug.
>
> I don't have the actual C89 or C99 standards in front of me at the moment,
> but this line in the C99 draft explains it:
>
> 6.7.8 Initialization
> [#2] No initializer shall attempt to provide a value for an
> object not contained within the entity being initialized.
I suspect that's talking about something different, e.g.:
int a[10] = { [20] = 0 };
It certainly can't be talking about expressions that contain side effects,
otherwise initializations like this wouldn't be allowed:
int c = getchar ();
> I believe it is a gcc extension that allows you to change the value of an
> unrelated variable inside an initializer.
I don't think so. I think that code is portable K&R C code, as well
as being portable C89 and C99 code, if the item being initialized is a
scalar.
Pre-C99 did have special rules for initializing aggregate types
(initializers had to be constants), and my guess is that the compiler
is confused and is thinking that Idx is a structure. Or maybe it
thinks Idx is 'long long' and that the 'long long' type is a
structure? Some earlier Sun C compilers had that bug.
