On Jan 15, 2009, at 9:05 AM, Douglas Gregor wrote:
>
> On Jan 14, 2009, at 11:41 PM, Chris Lattner wrote:
>
>> On Jan 14, 2009, at 3:24 PM, Douglas Gregor wrote:
>>>>> +/// @brief Determine the result of name lookup.
>>>>> +Sema::LookupResult::LookupKind Sema::LookupResult::getKind()
>>>>> const {
>>>>> + switch (StoredKind) {
>>>>> + case SingleDecl:
>>>>> + return (reinterpret_cast<Decl *>(First) != 0)? Found :
>>>>> NotFound;
>>>>> +
>>>>> + case OverloadedDeclFromIdResolver:
>>>>> + case OverloadedDeclFromDeclContext:
>>>>> + return FoundOverloaded;
>>>>> +
>>>>> + case AmbiguousLookup:
>>>>> + return Ambiguous;
>>>>> + }
>>>>> +
>>>>> + // We can't get here, but GCC complains nonetheless.
>>>>> + return Ambiguous;
>>>>> +}
>>>>>
>>>>
>>>> I believe our standard way of handling this is by having a default
>>>> case
>>>> that asserts.
>>>>
>>>> switch (StoredKind) {
>>>> default: assert(false && "Bad StoredKind");
>>>
>>> IIRC, I was getting bogus "function does not return a value"
>>> warnings
>>> from GCC when I've done this.
>>
>> Make sure the default case is at the start of the switch (Before
>> the first case) not at the end.
>
>
> My complaint about putting a default: in there still stands: it
> suppresses a useful warning, where we've omitted a case, and turns
> it into a run-time failure.
Does GCC produce a warning for cases like:
int foo() {
switch (someenum) {
case <all possible enums>:
return 0;
}
// no return
}
About falling off the end of the function?
-Chris
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
