[llvm-commits] [llvm] r133063 - in /llvm/trunk: include/llvm/CodeGen/ScheduleDAG.h lib/CodeGen/ScheduleDAG.cpp lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp test/CodeGen/X86/2011-06-14-PreschedRega

[Andrew Trick]

Can you see a project option titled LLVM_ENABLE_ASSERTIONS? Turning that OFF should disable the test as I explained. If that is ON then please verify that the compile line contains -UNDEBUG after any occurence of -DNDEBUG? If not, then that is a LLVM cmake bug that we can easily fix. I just need someone to verify it.

-Andy

On Jun 24, 2011, at 6:58 PM, Francois Pichet wrote:

> ok the error is:
> 1>  Command 0 Stderr:
> 1>  llc.EXE: Unknown command line argument '-stress-sched'.  Try:
> 'C:/dev/llvm/llvm_trunk2/bin/Release/llc.EXE -help'
> 1>  llc.EXE: Did you mean '-pre-RA-sched'?
> 
> I am always using latest svn.
> Seems like the ENABLE_ASSERTIONS thing is not working on MSVC.
> 
> 
> On Fri, Jun 24, 2011 at 9:36 PM, Andrew Trick <atrick at apple.com> wrote:
>> Running the tests with the release build is fine. You should not see any failures. This particular test should be skipped as unsupported.
>> 
>> If you're on r133725+, cmake should set the following line in lit "site config" (test/lit.site.cfg)
>> 
>> config.enable_assertions = 0
>> 
>> If you still have a problem, send me the details or file a bug. It should be easy to fix, the only issue here is that I don't have MSVC for testing.
>> 
>> You can disregard my "self-review" comment below. I disabled the test shortly after checking it in until I believed the build system was handling it.
>> 
>> -Andy
>> 
>> On Jun 24, 2011, at 6:17 PM, Francois Pichet wrote:
>> 
>>> This is sad, I recently took the habit of running the llvm tests at
>>> least once a day on Windows (MSVC 2010) and this test is the only one
>>> that constantly fails. I run them in in Release because a Debug run
>>> takes forever and there is no Release+Assert on Windows.
>>> 
>>> Just saying....
>>> 
>>> On Wed, Jun 15, 2011 at 1:56 PM, Andrew Trick <atrick at apple.com> wrote:
>>>> Self review: This test will fail in the no-Asserts build because -stress-sched is unavailable. It seems that llvm-lit always picks up the Asserts build. But is there some formal way to indicate that the test should not be run unless built with Asserts?
>>>> 
>>>> -Andy
>>>> 
>>>> On Jun 15, 2011, at 10:16 AM, Andrew Trick wrote:
>>>> 
>>>>> Author: atrick
>>>>> Date: Wed Jun 15 12:16:12 2011
>>>>> New Revision: 133063
>>>>> 
>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=133063&view=rev
>>>>> Log:
>>>>> Added -stress-sched flag in the Asserts build.
>>>>> Added a test case for handling physreg aliases during pre-RA-sched.
>>>>> 
>>>>> Added:
>>>>>   llvm/trunk/test/CodeGen/X86/2011-06-14-PreschedRegalias.ll
>>>>> Modified:
>>>>>   llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h
>>>>>   llvm/trunk/lib/CodeGen/ScheduleDAG.cpp
>>>>>   llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
>>>>>   llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
>>>>> 
>> 
>>