[llvm-commits] [llvm] r52528 - /llvm/trunk/test/FrontendC/2007-09-05-ConstCtor.c

[Dale Johannesen]

On Jun 23, 2008, at 10:36 AMPDT, Chris Lattner wrote:

>
> On Jun 23, 2008, at 10:27 AM, Dale Johannesen wrote:
>
>> I'm not very happy about this one.  'vcgt' is implementation-defined
>> and gcc properly warns about it;
>> but for historical reasons it's widely used inside Apple (widely
>> enough that the warning is suppressed),
>> so there really should be tests that exercise it.
>>
>> Obviously the best thing is to fix the test harness so it tolerates
>> warnings.
>
> The test didn't care about the use of the apple literal character
> thing, it cared about struct layout.

I know, that's not the point.  Regression test failures are caused by  
something
that has nothing to do with the original point of the test at least  
half the time IME.
We shouldn't be barred from using this construct in test cases.

>> On Jun 19, 2008, at 10:31 PMPDT, Chris Lattner wrote:
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> = 
>>> ====================================================================
>>> --- llvm/trunk/test/FrontendC/2007-09-05-ConstCtor.c (original)
>>> +++ llvm/trunk/test/FrontendC/2007-09-05-ConstCtor.c Fri Jun 20
>>> 00:31:04 2008
>>> @@ -1,7 +1,5 @@
>>> // RUN: %llvmgcc -xc -Os -c %s -o /dev/null
>>> // PR1641
>>> -// XFAIL: *
>>> -// See PR2452
>>>
>>> struct A {
>>> unsigned long l;
>>> @@ -10,7 +8,7 @@
>>> void bar(struct A *a);
>>>
>>> void bork() {
>>> -  const unsigned long vcgt = 'vcgt';
>>> +  const unsigned long vcgt = 1234;
>>> struct A a = { vcgt };
>>> bar(&a);
>>> }
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits