Jenkins asan+ubsan build failure / X86TargetLowering::LowerBUILD_VECTOR

[Cong Hou via llvm-commits]

Sorry for just noticing this problem and thanks for the fix, Kaelyn.
The fix looks good, and I think it is better to add an assertion
checking NumElems is no greater than 64.


thanks,
Cong


On Tue, Nov 24, 2015 at 3:02 PM, Kaelyn Takata <rikka at google.com> wrote:
> The following small patch fixes the asan error about the bit shift, but I'm
> not sure if it is the right thing to do or not. If it is, I'll commit it.
>
> On Tue, Nov 24, 2015 at 1:20 PM, Kaelyn Takata <rikka at google.com> wrote:
>>
>> I'm encountering the same error in test/CodeGen/X86/avg.ll when running
>> with asan enabled, due to a bit shift of 33 with a 32-bit integer.
>>
>> On Tue, Nov 24, 2015 at 10:56 AM, Matthias Braun via llvm-commits
>> <llvm-commits at lists.llvm.org> wrote:
>>>
>>> Hi,
>>>
>>> the greendragon asan+ubsan build started failing:
>>>
>>> http://lab.llvm.org:8080/green/job/clang-stage2-cmake-RgSan_check/782/
>>>
>>> I believe this was triggered by r253952 "[X86][SSE] Detect AVG pattern
>>> during instruction combine for SSE2/AVX2/AVX512BW."
>>>
>>> Looking at the failure this seems to be an cause by an existing bug in
>>> X86TargetLowering::LowerBUILD_VECTOR() which uses an unsigned variable as a
>>> bitset with a bit for each vector element. The test added contains some "64
>>> x XXX" vector types and overflows this.
>>>
>>> - Matthias
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>
>>
>