[llvm-testresults] evchen-ppc32 nightly tester results

[Bob Wilson]

Can you send me a bitcode file?

On May 5, 2010, at 5:28 PM, Dale Johannesen wrote:

> 
> On May 5, 2010, at 5:26 PMPDT, Evan Cheng wrote:
>> On May 5, 2010, at 5:17 PM, Dale Johannesen wrote:
>> On May 5, 2010, at 3:29 PMPDT, Dale Johannesen wrote:
>>> 
>>>> Building stepanov_vector crashes thus:
>>> 
>>> SSAUpdater is crashing because of this block, which exists on entry to GVN:
>>> 
>>> bb.i.i61.us:                                      ; preds = %bb.i.i61.us
>>> %404 = invoke i32 (i8*, ...)* @"\01_printf$LDBL128"(i8* getelementptr inbounds ([21 x i8]* @.str1, i32 0, i32 0), i32 0)
>>>        to label %bb.i.i61.us unwind label %lpad149 ; <i32> [#uses=0]
>>> 
>>> As the immediate dominator is 0 this loop crashes on the 2nd iteration:
>>> 
>>> 450 static bool IsDefInDomFrontier(const SSAUpdater::BBInfo *Pred,
>>> 451                                const SSAUpdater::BBInfo *IDom) {
>>> 452   for (; Pred != IDom; Pred = Pred->IDom) {
>>> 453     if (Pred->DefBB == Pred)
>>> 454       return true;
>>> 455   }
>>> 
>>> What is the expectation here?  Should SSAUpdater handle this, or should it have been removed earlier?
>>> 
>> Bob touched SSA updater last. :-)
> 
> 
> Last night, yeah, but it was broken before that.
> 
> I applied the obvious trivial fix to the loop above and get another crash later.  I don't think it's expecting unreachable blocks....
>