1.9 Next Steps

tonic · November 8, 2006, 6:39am

I created the 1.9 release branch last night. As a reminder, please do not check in any code changes to the release branch. Please send me email if you have changes that need to be merged into the release branch.

To check out the release branch:
cvs -d <CVS Repository> co -r release_19 llvm
cvs -d <CVS Repository> co -r release_19 llvm-test
cvs -d <CVS Repository> co -r release_19 llvm-gcc
svn co svn://anonsvn.opensource.apple.com/svn/llvm/tags/release-1.9 dst-directory

Unlike code changes, please go ahead and commit any documentation changes
to the release branch (and also mainline) directly without review. I am
hoping that all LLVMers could spare a few minutes to check over the
documents to make sure they are up to date. Please complete the documentation review by November 17th at 5PM PST.

I will be creating tarballs and binaries which will be available for
testing by November 13th.

Thanks again for all your hard work!

-Tanya

Emil_Mikulic1 · November 8, 2006, 5:55am

I'm getting a build error:

llvm[2]: Compiling llvmAsmParser.cpp for Release build
/rest/llvm/llvm/lib/AsmParser/llvmAsmParser.y: In function `int llvmAsmparse()':
/rest/llvm/llvm/lib/AsmParser/llvmAsmParser.y:2105: error: expected `;' before '}' token
/rest/llvm/llvm/lib/AsmParser/llvmAsmParser.y:2106: error: expected `;' before '}' token

with bison (GNU Bison) 1.75

This seems to fix it:

--- lib/AsmParser/llvmAsmParser.y.orig Wed Nov 8 16:53:24 2006
+++ lib/AsmParser/llvmAsmParser.y Wed Nov 8 16:53:34 2006
@@ -2102,8 +2102,8 @@
};

FnDeclareLinkage: /*default*/ |
- DLLIMPORT { CurFun.Linkage = GlobalValue::DLLImportLinkage } |
- EXTERN_WEAK { CurFun.Linkage = GlobalValue::DLLImportLinkage };
+ DLLIMPORT { CurFun.Linkage = GlobalValue::DLLImportLinkage; } |
+ EXTERN_WEAK { CurFun.Linkage = GlobalValue::DLLImportLinkage; };

FunctionProto : DECLARE { CurFun.isDeclare = true; } FnDeclareLinkage FunctionHeaderH {
$$ = CurFun.CurrentFunction;

Reid_Spencer · November 8, 2006, 6:20am

Hi Tanya,

I've been checking the state of the various llvm-test failures on
X86/Linux with GCC 3.4.6 and llvm-gcc4. I haven't finished this, but I
thought the following might be useful for other people that are testing
the release on Linux. Each group of failing tests below is followed by
a comment about why its failing.

llc /MultiSource/Applications/oggenc/oggenc
jit /MultiSource/Applications/oggenc/oggenc
cbe /MultiSource/Applications/oggenc/oggenc
^^ no idea what's causing these

jit /MultiSource/Applications/JM/ldecod/ldecod
^^ weird jit codegen problem produces wrong FP output, PR963

llc /MultiSource/Applications/JM/lencod/lencod
jit /MultiSource/Applications/JM/lencod/lencod
cbe /MultiSource/Applications/JM/lencod/lencod
^^ no idea

jit /MultiSource/Applications/obsequi/Obsequi
^^ no idea

cbe /MultiSource/Benchmarks/Prolangs-C/archie-client/archie
^^ cbe doesn't support inline asm (bzero lowering)

jit /MultiSource/Benchmarks/Prolangs-C/allroots/allroots
llc /MultiSource/Benchmarks/Prolangs-C/allroots/allroots
^^ no idea

llc /SingleSource/UnitTests/Vector/SSE/sse.stepfft
jit /SingleSource/UnitTests/Vector/SSE/sse.stepfft
llc /SingleSource/UnitTests/Vector/build2
jit /SingleSource/UnitTests/Vector/build2
cbe /SingleSource/UnitTests/Vector/build2
llc /SingleSource/UnitTests/Vector/build
jit /SingleSource/UnitTests/Vector/build
cbe /SingleSource/UnitTests/Vector/build
cbe /SingleSource/UnitTests/Vector/divides
llc /SingleSource/UnitTests/Vector/multiplies
jit /SingleSource/UnitTests/Vector/multiplies
cbe /SingleSource/UnitTests/Vector/multiplies
llc /SingleSource/UnitTests/Vector/simple
jit /SingleSource/UnitTests/Vector/simple
cbe /SingleSource/UnitTests/Vector/simple
llc /SingleSource/UnitTests/Vector/sumarray
jit /SingleSource/UnitTests/Vector/sumarray
cbe /SingleSource/UnitTests/Vector/sumarray
llc /SingleSource/UnitTests/Vector/sumarray-dbl
jit /SingleSource/UnitTests/Vector/sumarray-dbl
cbe /SingleSource/UnitTests/Vector/sumarray-dbl
^^ all of the above fail because my 3.4.6 compiler doesn't support
vectors well

llc /SingleSource/Regression/C++/EH/ctor_dtor_count-2
jit /SingleSource/Regression/C++/EH/ctor_dtor_count-2
cbe /SingleSource/Regression/C++/EH/ctor_dtor_count-2
llc /SingleSource/Regression/C++/EH/ctor_dtor_count
jit /SingleSource/Regression/C++/EH/ctor_dtor_count
cbe /SingleSource/Regression/C++/EH/ctor_dtor_count
llc /SingleSource/Regression/C++/EH/exception_spec_test
jit /SingleSource/Regression/C++/EH/exception_spec_test
cbe /SingleSource/Regression/C++/EH/exception_spec_test
llc /SingleSource/Regression/C++/EH/function_try_block
jit /SingleSource/Regression/C++/EH/function_try_block
cbe /SingleSource/Regression/C++/EH/function_try_block
llc /SingleSource/Regression/C++/EH/simple_rethrow
jit /SingleSource/Regression/C++/EH/simple_rethrow
cbe /SingleSource/Regression/C++/EH/simple_rethrow
llc /SingleSource/Regression/C++/EH/simple_throw
jit /SingleSource/Regression/C++/EH/simple_throw
cbe /SingleSource/Regression/C++/EH/simple_throw
llc /SingleSource/Regression/C++/EH/throw_rethrow_test
jit /SingleSource/Regression/C++/EH/throw_rethrow_test
cbe /SingleSource/Regression/C++/EH/throw_rethrow_test
llc /SingleSource/Regression/C++/BuiltinTypeInfo
cbe /SingleSource/Regression/C++/BuiltinTypeInfo
llc /SingleSource/Regression/C++/ofstream_ctor
cbe /SingleSource/Regression/C++/ofstream_ctor
^^ All of these fail due to C++ exception handling not implemented in
llvm-gcc4

llc /SingleSource/Benchmarks/CoyoteBench/fftbench
jit /SingleSource/Benchmarks/CoyoteBench/fftbench
^^ PR990

llc /SingleSource/Benchmarks/Shootout-C++/except
jit /SingleSource/Benchmarks/Shootout-C++/except
cbe /SingleSource/Benchmarks/Shootout-C++/except
llc /SingleSource/Benchmarks/Misc-C++/bigfib
jit /SingleSource/Benchmarks/Misc-C++/bigfib
^^ C++ EH again

llc /SingleSource/CustomChecked/oopack_v1p8
cbe /SingleSource/CustomChecked/oopack_v1p8
^^ Not sure, but I suspect its lack of vector support again.

Chris_Lattner4 · November 8, 2006, 7:07am

Definitely a bug, thanks!

Tanya, please pull these patches into the release when convenient:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20061106/039618.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20061106/039619.html

-Chris

tonic · November 8, 2006, 6:15pm

Thanks Reid. Most of these are probably not regressions but I will verify and file bugzilla bugs for them once I finish all my pre-release testing.

-Tanya

Evan_Cheng1 · November 10, 2006, 9:48pm

Hi Tanya,

Please apply these patches to the release branch:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20061106/039776.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20061106/039781.html

Thanks,

Evan

tonic · November 10, 2006, 11:54pm

Please apply these patches to the release branch:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-
Mon-20061106/039776.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-
Mon-20061106/039781.html

I think I have these merged in correctly. I will verify that the bug is fixed later tonight.

Thanks!
-Tanya