[llvm-testresults] buildbot failure in smooshlab on clang-x86_64-darwin10-selfhost-rel
daniel_dunbar at apple.com
daniel_dunbar at apple.com
Fri May 14 15:23:42 PDT 2010
The Buildbot has detected a new failure of clang-x86_64-darwin10-selfhost-rel on smooshlab.
Full details are available at:
http://smooshlab.apple.com:8010/builders/clang-x86_64-darwin10-selfhost-rel/builds/2071
Buildbot URL: http://smooshlab.apple.com:8010/
Buildslave for this Build: smoosh-02
Build Reason:
Build Source Stamp: 103819
Blamelist: bwilson,dgregor,fjahanian,kremenek,void
BUILD FAILED: failed test-llvm test-llvm_1
sincerely,
-The Buildbot
================================================================================
CHANGES:
Files:
gcc/config.gcc
gcc/config/arm/arm-protos.h
gcc/config/arm/arm.c
gcc/config/arm/arm_neon.h
gcc/config/arm/arm_neon_gcc.h
gcc/config/arm/arm_neon_std.h
gcc/config/arm/neon-gen-std.ml
gcc/config/arm/neon-gen.ml
At: Fri 14 May 2010 14:34:06
Changed By: bwilson
Comments: llvm-gcc's implementation of Neon types and intrinsics, which follows ARM's
specifications, has been causing problems for porting code from gcc. GCC's
implementation defines the Neon types as plain vector types so that you can
freely intermix builtin vector operators with Neon intrinsics. The ARM
standard Neon types are "containerized vectors", i.e., structs, so you cannot
use them with the usual vector operators (+, -, *, etc.). This change adds
an optional gcc-compatibility mode for the Neon types and intrinsics to
support code written in that style of intermixing intrinsics and operators.
Before going into details, I also want to mention that the plan is for Clang
to define overloaded intrinsics that will accept either ARM's containerized
types or plain vector types. That is not yet implemented, but once it is
available, Clang users can make a choice of declaring Neon vectors with either
kind of type. If they go with the standard containerized vector types,
they'll get strict conformance to ARM's specifications but will not be able
to use vector operators with those values. If they go with plain vector
types, they'll lose the standard conformance but be able to mix-and-match
intrinsics with operators.
This patch is intended as a step in the direction of the Clang solution, but
since llvm-gcc does not support function overloading in C (besides the fact
that it implements the Neon intrinsics as preprocessor macros for other
reasons), llvm-gcc users will have to consistently use one style or the other.
If they define ARM_NEON_GCC_COMPATIBILITY before including <arm_neon.h>,
then the Neon intrinsics will be defined to work on plain vector types.
Otherwise, they'll continue to get the standard versions.
The patch brings back some of the awful arm_mangle_types code that fakes
the C++ mangling for Neon types to make them be mangled as if they were
the containerized vector structs. I've addressed the previous problems
with that code by defining the Neon vector types as built-in types so that
the mangler can recognize them by their unique type nodes. The motivation
for resurrecting that code is that we don't want binary compatibility problems
between llvm-gcc with ARM_NEON_GCC_COMPATIBILITY and Clang. However,
the expected use of the ARM_NEON_GCC_COMPATIBILITY mode is that variables
will be declared with plain vector types. That usage should continue to
work without changes in Clang (where ARM_NEON_GCC_COMPATIBILITY will not
be needed). Likewise, code that does not use that mode will continue to
work unmodified in Clang. Code using the standard Neon type names when
ARM_NEON_GCC_COMPATIBILITY is defined may require changes when moving to
Clang, but those are basically the same changes that would otherwise be
required when porting from gcc to llvm-gcc.
Down to the details.... There are now 2 versions of the neon-gen.ml
generator for arm_neon.h, and 2 versions of the arm_neon.h output. The
neon-gen-std.ml version generates arm_neon_std.h, which is the version using
the standard containerized vector types. The neon-gen.ml version generates
arm_neon_gcc.h, which is the gcc-compatible version. Neither of those headers
should ever be used directly, since they will not be available with clang.
The arm_neon.h header is now a simple wrapper that selects between the 2
versions based on whether ARM_NEON_GCC_COMPATIBILITY is defined.
Properties:
Files:
lib/Target/TargetMachine.cpp
lib/Target/X86/X86RegisterInfo.cpp
At: Fri 14 May 2010 14:34:06
Changed By: void
Comments: Revert r103804. The comment is correct.
Properties:
Files:
include/clang-c/Index.h
test/Index/print-typekind.c
tools/c-index-test/c-index-test.c
tools/libclang/CMakeLists.txt
tools/libclang/CXTypes.cpp
tools/libclang/libclang.darwin.exports
tools/libclang/libclang.exports
At: Fri 14 May 2010 14:34:06
Changed By: kremenek
Comments: Add CXType and an initial set of supporting functions to libclang. This exposes details of
Clang's representation of the C type system to clients. It is nowhere near complete, and will
be expanded on demand.Properties:
Files:
lib/CodeGen/CGExprAgg.cpp
test/CodeGenCXX/dynamic-cast.cpp
At: Fri 14 May 2010 14:34:06
Changed By: dgregor
Comments: Emit an lvalue dynamic_cast even if the result is not used. Another
part (or possibly all) of PR7132.
Properties:
Files:
lib/Sema/SemaDecl.cpp
test/Parser/cxx-undeclared-identifier.cpp
At: Fri 14 May 2010 14:39:21
Changed By: fjahanian
Comments: Patch to fix a crash on incomplete class declaration.
Radar 7923673.
Properties:
File: lib/Target/X86/X86RegisterInfo.cpp
At: Fri 14 May 2010 14:44:01
Changed By: void
Comments: This should happen if there are no calls, not if it just doesn't adjust the
stack.
Properties:
File: test/CodeGenCXX/dynamic-cast.cpp
At: Fri 14 May 2010 14:54:02
Changed By: dgregor
Comments: Tweak test so that it does not require <typeinfo>Properties:
LOGS:
Last 10 lines of 'stdio':
LLVM :: CodeGen/X86/2009-12-11-TLSNoRedZone.ll
LLVM :: CodeGen/X86/tailcall-largecode.ll
LLVM :: CodeGen/X86/tailcallstack64.ll
Expected Passes : 4714
Expected Failures : 27
Unsupported Tests : 523
Unexpected Failures: 3
make[1]: *** [check-local-lit] Error 1
make: *** [check-lit] Error 2
Last 10 lines of 'fail':
LLVM :: CodeGen/X86/2009-12-11-TLSNoRedZone.ll
LLVM :: CodeGen/X86/tailcall-largecode.ll
LLVM :: CodeGen/X86/tailcallstack64.ll
Last 10 lines of 'xfail':
LLVM :: MC/AsmParser/X86/x86_32-bit_cat.s
LLVM :: MC/AsmParser/X86/x86_32-mismatched-add.s
LLVM :: MC/AsmParser/X86/x86_32-encoding.s
LLVM :: MC/AsmParser/exprs-invalid.s
LLVM :: MC/AsmParser/directive_lsym.s
LLVM :: Transforms/DeadArgElim/deadexternal.ll
LLVM :: Transforms/GVN/rle-no-phi-translate.ll
LLVM :: Transforms/IndVarSimplify/loop_evaluate_6.ll
LLVM :: Transforms/TailCallElim/nocapture.ll
LLVM :: Transforms/TailDup/2008-06-11-AvoidDupLoopHeader.ll
Last 10 lines of 'unsupported':
LLVM :: LLVMC/C++/hello.cpp
LLVM :: LLVMC/C++/together.cpp
LLVM :: LLVMC/C/sink.c
LLVM :: LLVMC/C/emit-llvm.c
LLVM :: LLVMC/C/hello.c
LLVM :: LLVMC/C/include.c
LLVM :: LLVMC/C/opt-test.c
LLVM :: LLVMC/C/wall.c
LLVM :: LLVMC/ObjC++/hello.mm
LLVM :: LLVMC/ObjC/hello.m
Last 10 lines of '2009-12-11-tlsnoredzone.ll':
<stdin>:11:13: error: CHECK-NOT: string occurred!
movq %rax, -8(%rsp)
^
/Users/buildslave/zorg/buildbot/smooshlab/slave/build.clang-x86_64-darwin10-selfhost-rel/llvm.src/test/CodeGen/X86/2009-12-11-TLSNoRedZone.ll:22:22: note: CHECK-NOT: pattern specified here
; CHECK-NOT: -8(%rsp)
^
--
********************
Last 10 lines of 'tailcall-largecode.ll':
<stdin>:11:1: note: scanning from here
.Ltmp0:
^
<stdin>:22:4: note: possible intended match here
movl $7, 8(%rsp)
^
--
********************
Last 10 lines of 'tailcallstack64.ll':
/Users/buildslave/zorg/buildbot/smooshlab/slave/build.clang-x86_64-darwin10-selfhost-rel/llvm.src/test/CodeGen/X86/tailcallstack64.ll:7:10: error: expected string not found in input
; CHECK: movl 40(%rsp), %r10d
^
<stdin>:7:2: note: scanning from here
movl 16(%rsp), %r10d
^
--
********************
Last 10 lines of 'stdio':
LLVM :: CodeGen/X86/2009-12-11-TLSNoRedZone.ll
LLVM :: CodeGen/X86/tailcall-largecode.ll
LLVM :: CodeGen/X86/tailcallstack64.ll
Expected Passes : 4714
Expected Failures : 27
Unsupported Tests : 523
Unexpected Failures: 3
make[1]: *** [check-local-lit] Error 1
make: *** [check-lit] Error 2
Last 10 lines of 'fail':
LLVM :: CodeGen/X86/2009-12-11-TLSNoRedZone.ll
LLVM :: CodeGen/X86/tailcall-largecode.ll
LLVM :: CodeGen/X86/tailcallstack64.ll
Last 10 lines of 'xfail':
LLVM :: MC/AsmParser/X86/x86_32-bit_cat.s
LLVM :: MC/AsmParser/X86/x86_32-encoding.s
LLVM :: MC/AsmParser/X86/x86_32-bit.s
LLVM :: MC/AsmParser/directive_lsym.s
LLVM :: MC/AsmParser/exprs-invalid.s
LLVM :: Transforms/DeadArgElim/deadexternal.ll
LLVM :: Transforms/GVN/rle-no-phi-translate.ll
LLVM :: Transforms/IndVarSimplify/loop_evaluate_6.ll
LLVM :: Transforms/TailCallElim/nocapture.ll
LLVM :: Transforms/TailDup/2008-06-11-AvoidDupLoopHeader.ll
Last 10 lines of 'unsupported':
LLVM :: LLVMC/C++/hello.cpp
LLVM :: LLVMC/C++/together.cpp
LLVM :: LLVMC/C/emit-llvm.c
LLVM :: LLVMC/C/hello.c
LLVM :: LLVMC/C/include.c
LLVM :: LLVMC/C/opt-test.c
LLVM :: LLVMC/C/sink.c
LLVM :: LLVMC/C/wall.c
LLVM :: LLVMC/ObjC++/hello.mm
LLVM :: LLVMC/ObjC/hello.m
Last 10 lines of '2009-12-11-tlsnoredzone.ll':
<stdin>:11:13: error: CHECK-NOT: string occurred!
movq %rax, -8(%rsp)
^
/Users/buildslave/zorg/buildbot/smooshlab/slave/build.clang-x86_64-darwin10-selfhost-rel/llvm.src/test/CodeGen/X86/2009-12-11-TLSNoRedZone.ll:22:22: note: CHECK-NOT: pattern specified here
; CHECK-NOT: -8(%rsp)
^
--
********************
Last 10 lines of 'tailcall-largecode.ll':
<stdin>:11:1: note: scanning from here
.Ltmp0:
^
<stdin>:22:4: note: possible intended match here
movl $7, 8(%rsp)
^
--
********************
Last 10 lines of 'tailcallstack64.ll':
/Users/buildslave/zorg/buildbot/smooshlab/slave/build.clang-x86_64-darwin10-selfhost-rel/llvm.src/test/CodeGen/X86/tailcallstack64.ll:7:10: error: expected string not found in input
; CHECK: movl 40(%rsp), %r10d
^
<stdin>:7:2: note: scanning from here
movl 16(%rsp), %r10d
^
--
********************
More information about the llvm-testresults
mailing list