[release_29] Good status of ppc-redhat-linux on Fedora 12 PS3

Good morning.

LLVM and clang can be built successfully on Fedora 12 PS3.

On RC1, only one test failed.
test/CodeGen/X86/fold-pcmpeqd-0.ll

On release_29 branch, all llvm tests can pass.
(I don't mention clang tests :stuck_out_tongue: )

...Takumi

Fedora release 12 (Constantine)
Linux speedking.localdomain 2.6.32.23-170.fc12.ppc64 #1 SMP Mon Sep 27
17:09:35 UTC 2010 ppc64 ppc64 ppc64 GNU/Linux

llvm config.status 2.9svn
configured by ../../llvm/configure, generated by GNU Autoconf 2.60,
  with options "'-C' '--build=ppc-redhat-linux' '--enable-targets=all'
'--enable-optimized' 'build_alias=ppc-redhat-linux'
'--with-optimize-option=-O3 -Werror'
'--prefix=/home/chapuni/BUILD/llvm-ppc-static/install/stage1'"

Failing Tests (39):
    Clang :: CXX/expr/expr.unary/expr.unary.noexcept/cg.cpp
    Clang :: CodeGen/bitfield-promote.c
    Clang :: CodeGenCXX/debug-info-byval.cpp
    Clang :: CodeGenCXX/debug-info-namespace.cpp
    Clang :: CodeGenCXX/vtable-debug-info.cpp
    Clang :: Driver/hello.c
    Clang :: Index/c-index-api-loadTU-test.m
    Clang :: Index/c-index-getCursor-test.m
    Clang :: Index/c-index-pch.c
    Clang :: PCH/chain-cxx.cpp
    Clang :: PCH/chain-remap-types.m
    Clang :: PCH/chain-selectors.m
    Clang :: PCH/check-deserializations.cpp
    Clang :: PCH/cuda-kernel-call.cu
    Clang :: PCH/cxx-friends.cpp
    Clang :: PCH/cxx-namespaces.cpp
    Clang :: PCH/cxx-static_assert.cpp
    Clang :: PCH/cxx-templates.cpp
    Clang :: PCH/cxx-traits.cpp
    Clang :: PCH/cxx-typeid.cpp
    Clang :: PCH/cxx-using.cpp
    Clang :: PCH/cxx-variadic-templates.cpp
    Clang :: PCH/cxx_exprs.cpp
    Clang :: PCH/exprs.c
    Clang :: PCH/headersearch.cpp
    Clang :: PCH/missing-file.cpp
    Clang :: PCH/namespaces.cpp
    Clang :: PCH/objc_import.m
    Clang :: PCH/objc_methods.m
    Clang :: PCH/objc_property.m
    Clang :: PCH/objcxx-ivar-class.mm
    Clang :: PCH/pragma-diag-section.cpp
    Clang :: PCH/reinclude.cpp
    Clang :: PCH/struct.c
    Clang :: PCH/typo.m
    Clang :: PCH/va_arg.cpp
    Clang :: Sema/stdcall-fastcall.c
    Clang :: Sema/x86-builtin-palignr.c
    Clang :: SemaCXX/attr-regparm.cpp

  Expected Passes : 8116
  Expected Failures : 68
  Unsupported Tests : 542
  Unexpected Failures: 39

Good morning.

Hi Nakamura,

LLVM and clang can be built successfully on Fedora 12 PS3.

Hooray! :slight_smile:

On RC1, only one test failed.
test/CodeGen/X86/fold-pcmpeqd-0.ll

Eric commented that this should be fixed on the release branch right now.

On release_29 branch, all llvm tests can pass.

Woo! :slight_smile:

(I don't mention clang tests :stuck_out_tongue: )

.>

...Takumi

Fedora release 12 (Constantine)
Linux speedking.localdomain 2.6.32.23-170.fc12.ppc64 #1 SMP Mon Sep 27
17:09:35 UTC 2010 ppc64 ppc64 ppc64 GNU/Linux

llvm config.status 2.9svn
configured by ../../llvm/configure, generated by GNU Autoconf 2.60,
with options "'-C' '--build=ppc-redhat-linux' '--enable-targets=all'
'--enable-optimized' 'build_alias=ppc-redhat-linux'
'--with-optimize-option=-O3 -Werror'
'--prefix=/home/chapuni/BUILD/llvm-ppc-static/install/stage1'"

Failing Tests (39):

Hrm. Could you work with the Clang guys to prioritize these failures? If they're "real", we need to get them fixed soon.

Thank you for the testing!

-bw

Good morning, Bill!

On RC1, only one test failed.
test/CodeGen/X86/fold-pcmpeqd-0.ll

Eric commented that this should be fixed on the release branch right now.

Of course I knew! It was the reason why I checked also on release_29.

Failing Tests (39):

Hrm. Could you work with the Clang guys to prioritize these failures? If they're "real", we need to get them fixed soon.

I have given up to investigate. As you know, PS3,Cell BE PPU, is too slow.
And I heard clang on ppc might be inmature.
I will post clang test log, if I can, ... unless I got power cut!
Good day!

...Takumi

Failing Tests (39):

Hrm. Could you work with the Clang guys to prioritize these failures? If they're "real", we need to get them fixed soon.

Bill and guys, here is a test log on ppc-redhat-linux, though not so
small, excuse me!

...Takumi

ppc-redhat-linux.log (167 KB)

Bill and guys, here is a gzip-ed test log on ppc-redhat-linux, though
not so small, excuse me!
...Takumi

ppc-redhat-linux.log.gz (17.8 KB)

CodeGenCXX/debug-info-byval.cpp is a combination of a badly written
test and what appears to be a crash-on-valid in the PPC backend.

CodeGenCXX/vtable-debug-info.cpp appears to be a combination of a
badly written test and the PPC backend generating invalid assembly.

I have no idea what the Driver/hello.c issue is; it's essentially
checking that clang can compile a runnable "Hello World" program.

Index/c-index-api-loadTU-test.m and Index/c-index-getCursor-test.m
appear to be a crash in the indexing code.

Sema/stdcall-fastcall.c, Sema/x86-builtin-palignr.c, and
SemaCXX/attr-regparm.cpp, and CodeGen/bitfield-promote.c are simply
buggy tests, as far as I can tell.

I'm not sure what's going on with CodeGenCXX/debug-info-namespace.cpp.

All of the other failures appear to be a bug in the PCH
reading/writing code; probably something isn't correctly handling
big-endian architectures.

-Eli

Good morning.

Hi Nakamura,

LLVM and clang can be built successfully on Fedora 12 PS3.

Hooray! :slight_smile:

On RC1, only one test failed.
test/CodeGen/X86/fold-pcmpeqd-0.ll

Eric commented that this should be fixed on the release branch right now.

On release_29 branch, all llvm tests can pass.

Woo! :slight_smile:

(I don't mention clang tests :stuck_out_tongue: )

.>

...Takumi

Fedora release 12 (Constantine)
Linux speedking.localdomain 2.6.32.23-170.fc12.ppc64 #1 SMP Mon Sep 27
17:09:35 UTC 2010 ppc64 ppc64 ppc64 GNU/Linux

llvm config.status 2.9svn
configured by ../../llvm/configure, generated by GNU Autoconf 2.60,
with options "'-C' '--build=ppc-redhat-linux' '--enable-targets=all'
'--enable-optimized' 'build_alias=ppc-redhat-linux'
'--with-optimize-option=-O3 -Werror'
'--prefix=/home/chapuni/BUILD/llvm-ppc-static/install/stage1'"

Failing Tests (39):

Hrm. Could you work with the Clang guys to prioritize these failures? If they're "real", we need to get them fixed soon.

There do appear to be "real" bugs in both the indexing code and the
PCH code. The rest of the failures appear to be a combination of
tests which are outright wrong and tests which make questionable
assumptions about the host.

-Eli