LLVM 4.0.1-rc1 has been tagged

Hi,

I'm seeming new regressions form 4.0.0 for mips big endian:

    DataFlowSanitizer-mips64 :: custom.cc
    DataFlowSanitizer-mips64 :: propagate.c
    SanitizerCommon-asan-mips-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-asan-mips-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-asan-mips64-Linux :: Linux/getpwnam_r_invalid_user.cc
    SanitizerCommon-asan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-asan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-lsan-mips64-Linux :: Linux/getpwnam_r_invalid_user.cc
    SanitizerCommon-lsan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-lsan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-msan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-msan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc

I'm already aware of the dataflow sanitizer bug. The sanitizer_coverage_trace_pc_guard(-dso) failures for
(a, l, m)san-mips64-Linux is an occurrence of a existing failure.

Little endian and debian are ok.

I've uploaded the binaries.

SHA1(clang+llvm-4.0.1-rc1-mips-linux-gnu.tar.xz)= bd5dd8ccc1e584c6846dea2b743a4227506a8f6b
SHA1(clang+llvm-4.0.1-rc1-mipsel-linux-gnu.tar.xz)= 4cc5ad7f6ab763a8b93a3168ed6420e1db4b8052
SHA1(clang+llvm-4.0.1-rc1-x86_64-linux-gnu-debian8.tar.xz)= bac30cd1820e0a3e8abd45e5ca836ea3ef0f78dc

Thanks,
Simon

Hi,

I'm seeming new regressions form 4.0.0 for mips big endian:

    DataFlowSanitizer-mips64 :: custom.cc
    DataFlowSanitizer-mips64 :: propagate.c
    SanitizerCommon-asan-mips-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-asan-mips-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-asan-mips64-Linux :: Linux/getpwnam_r_invalid_user.cc
    SanitizerCommon-asan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-asan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-lsan-mips64-Linux :: Linux/getpwnam_r_invalid_user.cc
    SanitizerCommon-lsan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-lsan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-msan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-msan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc

Can you bisect to see which commit broke these tests?

-Tom

Hi,

I'm seeming new regressions form 4.0.0 for mips big endian:

    DataFlowSanitizer-mips64 :: custom.cc
    DataFlowSanitizer-mips64 :: propagate.c
    SanitizerCommon-asan-mips-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-asan-mips-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-asan-mips64-Linux :: Linux/getpwnam_r_invalid_user.cc
    SanitizerCommon-asan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-asan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-lsan-mips64-Linux :: Linux/getpwnam_r_invalid_user.cc
    SanitizerCommon-lsan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-lsan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc
    SanitizerCommon-msan-mips64-Linux :: sanitizer_coverage_trace_pc_guard-dso.cc
    SanitizerCommon-msan-mips64-Linux :: sanitizer_coverage_trace_pc_guard.cc

Can you bisect to see which commit broke these tests?

Did you ever figure out what caused the regressions? Do you think these are
release blockers?

-Tom

I had no luck tracking down the commit that cause Compiler-rt to change it's supported
targets for the machine I was building on. It may be due to change in the machine's
configuration.

I'm not considering them release blockers as some are known defects partially caused by
the test machine's version of libc. The Dataflow sanitizer regression I have a posted patch
for but it hasn't gone through review yet. But the underlying bug was there for some time,
it wasn't tested.

Thanks,
Simon