[llvm-commits] [llvm-gcc-4.2] r48440 - in /llvm-gcc-4.2/trunk/gcc: config/rs6000/llvm-rs6000.cpp config/rs6000/rs6000.h llvm-abi.h

[Dale Johannesen]

On Mar 17, 2008, at 12:18 AM, Duncan Sands wrote:

> Hi Dale,
>
>> -               !aggregate_value_p(type, current_function_decl) &&
>> -               // FIXME: this is a hack around returning 'complex  
>> double' by-val
>> -               // which returns in r3/r4/r5/r6 on PowerPC.
>> -               TREE_INT_CST_LOW(TYPE_SIZE_UNIT(type)) <= 8) {
>> +               !aggregate_value_p(type, current_function_decl)) {
>
> while you are there, how about removing this bogus use of  
> current_function_decl?
> After all, when analyzing a call it is not the current function you  
> want but the
> callee...

That's an interesting point.  One of the 4 places this is called is  
from StartFunctionBody and current_function_decl is OK in that case.   
The other 3 do look wrong, but it's not obvious why this isn't causing  
errors if it is.   I guess I'll change it and see what breaks....