llvm dev - Mar 2011 - [LLVMdev] [cfe-dev] Announcing: LLVM 2.9 RC2 Testing Phase

On Fri, Mar 25, 2011 at 15:07, Bill Wendling <wendling at apple.com>
wrote:
> Hi all,
>
> Well! we had a rather fruitful phase 1 testing round. Several issues were
> addressed. After a bit of a delay, we are ready for phase 2 testing.
>
> This phase is to make sure that no patches submitted to fix problems and
> complete features in phase 1 caused further difficulties. Please download
> the sources, compile them, and then compile as much as you can with both
> clang and llvm-gcc.
>
> Sources:
>
> clang-2.9rc2.src.tar.gz
> llvm-2.9rc2.src.tar.gz
> llvm-gcc-4.2-2.9rc2.src.tar.gz
>  llvm-test-2.9rc2.src.tar.gz
>
> Available here:
>
> http://llvm.org/pre-releases/2.9/
>
> -bw
>
> *We are no longer accepting patches to complete features.* Any unfinished
> features should have been disabled all ready. The only patches that we will
> accept are those that fix a regression from the 2.8 release. We were fairly
> flexible with allowing people to commit to the release branch in phase 1.
> But now only the release manager may commit to the release branch. And
> the patches *must* be approved by the code owners. No exceptions.
>
Tested with gcc 4.5.3 (20100324) on i686-w64-windows:

http://i18n-zh.googlecode.com/files/gcc-4.5.3-windows_20110324.tar.xz
http://i18n-zh.googlecode.com/files/llvm-clang-2.9-rc2-win32.tar.xz
http://i18n-zh.googlecode.com/files/llvm-clang-2.9-rc2-win32.zip

[[[
********************
Testing: 0 .. 10.. 20.. 30.. 40.. 50.. 60.. 70.. 80.. 90..
Testing Time: 652.86s
********************
Failing Tests (1):
    LLVM :: CodeGen/X86/h-registers-1.ll

  Expected Passes    : 5246
  Expected Failures  : 46
  Unsupported Tests  : 546
  Unexpected Failures: 1
]]]

[[[
Failing Tests (1):
    LLVM :: CodeGen/X86/h-registers-1.ll

  Expected Passes    : 5246
  Expected Failures  : 46
  Unsupported Tests  : 546
  Unexpected Failures: 1

FAIL: LLVM :: CodeGen/X86/h-registers-1.ll (2430 of 5839)
******************** TEST 'LLVM :: CodeGen/X86/h-registers-1.ll' FAILED
********
************
Script:
--
C:/var/tmp/llvm-clang-2.9-rc2-obj/bin/./llc.EXE <
C:/var/tmp/llvm-clang-2.9-rc2/test/CodeGen/X86/h-registers-1.ll
-march=x86-64 > C:/var/tmp/llvm-clang-2.9-rc2-
obj/test/CodeGen/X86/Output/h-registers-1.ll.tmp
grep {movzbl    %\[abcd\]h,}
C:/var/tmp/llvm-clang-2.9-rc2-obj/test/CodeGen/X86/Output/h-registers-1.ll.tmp
| C:/var/tmp/llvm-clang-2.9-rc2-obj/bin/./count 8
grep {%\[abcd\]h}
C:/var/tmp/llvm-clang-2.9-rc2-obj/test/CodeGen/X86/Output/h-registers-1.ll.tmp
| C:/var/tmp/llvm-clang-2.9-rc2-obj/bin/./not grep {%r\[\[:digit:\]\]*d}
--
Exit Code: 1
Command Output (stderr):
--
Expected 8 lines, got 7.
--
]]]

--
Dongsheng
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20110325/9804ec93/attachment.html>

Good morning, Bill.

On Sat, Mar 26, 2011 at 5:12 AM, Bill Wendling <wendling at apple.com>
wrote:
>> Failing Tests (1):
>>     LLVM :: CodeGen/X86/h-registers-1.ll
> Could you send the output for this command? I haven't seen this failure
crop up before...
I missed it (and it was not in RC1). IIRC, Jakob's commit would affect
it on Windows.
It has been fixed on ToT, in r127800.

...Takumi