On Thu, Feb 7, 2013 at 11:29 AM, Jordan Rose <[email protected]> wrote:
>
> On Feb 7, 2013, at 11:02 , Chad Rosier <[email protected]> wrote:
>
>>
>> On Feb 7, 2013, at 10:57 AM, David Blaikie <[email protected]> wrote:
>>
>>> On Thu, Feb 7, 2013 at 10:49 AM, Chad Rosier <[email protected]> wrote:
>>>>
>>>> On Feb 7, 2013, at 10:39 AM, David Blaikie <[email protected]> wrote:
>>>>
>>>>> On Thu, Feb 7, 2013 at 10:32 AM, Chad Rosier <[email protected]> wrote:
>>>>>> Author: mcrosier
>>>>>> Date: Thu Feb  7 12:32:25 2013
>>>>>> New Revision: 174640
>>>>>>
>>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=174640&view=rev
>>>>>> Log:
>>>>>> Testcase for r174477.
>>>>>>
>>>>>> Added:
>>>>>>  cfe/trunk/test/Sema/invalid-cast.cpp
>>>>>>
>>>>>> Added: cfe/trunk/test/Sema/invalid-cast.cpp
>>>>>> URL: 
>>>>>> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Sema/invalid-cast.cpp?rev=174640&view=auto
>>>>>> ==============================================================================
>>>>>> --- cfe/trunk/test/Sema/invalid-cast.cpp (added)
>>>>>> +++ cfe/trunk/test/Sema/invalid-cast.cpp Thu Feb  7 12:32:25 2013
>>>>>> @@ -0,0 +1,11 @@
>>>>>> +// RUN: %clang_cc1 -verify -fsyntax-only %s
>>>>>> +// expected-no-diagnostics
>>>>>> +// <rdar://problem/13153516> - This previously triggered an assertion 
>>>>>> failure.
>>>>>
>>>>> Requires: Asserts ?
>>>>
>>>> I saw similar test cases that were checking things that perviously 
>>>> asserted, but didn't require an assert build.  I don't think it is 
>>>> necessary, but I also don't object to adding it.
>>>
>>> Yeah, no actually strong opinion either - not sure if it's cheaper for
>>> the infrastructure to check the requirement & not run the test than to
>>> just run it anyway... probably slightly cheaper to exclude it on
>>> non-assert builds, but *shrug*
>>
>> Haha.. Ok, I'll just leave it as is.  Thanks, David.
>
> FWIW I'd prefer to have assertion failure tests still run in -Asserts builds. 
> A crash is also a useful build result. (This is the sort of thing that 
> happens when useful work is accidentally done within the call to assert().)

Sure, but there's no guarantee that violating the invariant would lead
to a crash/test failure. The idea is that running this test on a
non-assert build doesn't necessarily provide confidence that the bug
is fixed. Adding "Requires: asserts" documents the fact that this test
is only intended to catch the bug under an asserts build.

A little bit "test theory" rather than terribly important/pragmatic, though.

- David
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to