[llvm-testresults] buildbot failure in llvm on llvm-gcc-x86_64-darwin10-cross-i686-linux

[Eric Christopher]

On Feb 5, 2011, at 2:10 PM, Bill Wendling wrote:

> On Feb 4, 2011, at 6:48 PM, Jakob Stoklund Olesen wrote:
> 
>> On Feb 4, 2011, at 5:44 PM, Eric Christopher wrote:
>> 
>>> ../../../../llvm-gcc.src/libstdc++-v3/libsupc++/vec.cc:359: internal compiler error: Abort trap
>>> 
>>> CallSiteBase(CallTy *CI) : I(CI, true) { assert(CI); }
>>> 
>>> Aroo?
>>> 
>>>  if (CallInst *CI = dyn_cast<CallInst>(U)) {
>>>    // Turning an indirect call into a direct call is a BIG win
>>>    if (CI->getCalledValue() == V)
>>>      Bonus += ConstantFunctionBonus(CallSite(CI), C);
>>>  } else if (InvokeInst *II = dyn_cast<InvokeInst>(U)) {
>>>    // Turning an indirect call into a direct call is a BIG win
>>>    if (II->getCalledValue() == V)
>>>      Bonus += ConstantFunctionBonus(CallSite(CI), C);
>>>  }
>>> 
>>> when is CI going to be NULL here?
>> 
>> The else branch is using CI where it will always be null.
>> 
>> ISTR that clang would warn about such uses. Maybe that was disabled.
>> 
> Is CI actually accessible in the else block? I thought that the scope of a variable defined in a conditional statement was the immediate scope afterwards... So only the "true" part of this if-then-else statement...

It is, lamely enough, I've run into this a couple of times. Fun huh?

-eric