[cfe-dev] ConstantArrayType's size is not always of pointer width

[Sebastian Redl]

Eli Friedman wrote:
> On Tue, Dec 2, 2008 at 9:03 AM, Sebastian Redl
> <sebastian.redl at getdesigned.at> wrote:
>   
>> Ah, OK. But is using the width of the constant the right thing to do?
>> Shouldn't an array size ultimately be of type size_t?
>>     
>
> It'll ultimately be allocated that way, but casting doesn't really
> seem necessary; we definitely don't want to truncate the size.
>   
So clients of the AST simply must not make assumptions about the width 
of array size and array index expressions. That's fine with me, but I've 
already fixed three 64-bit bugs in the static verifier that were caused 
by this assumption, so perhaps this should be documented very clearly 
somewhere.

Sebastian