[10.0.0 Release] Release Candidate 2 is here

Project Infrastructure Release Testers
1

Hello everyone,

Release Candidate 2 was tagged earlier today as llvmorg-10.0.0-rc2. It
includes 98 commits since the previous release candidate.

Source code and docs are available at
LLVM 10.0.0 Release Candidates and
https://github.com/llvm/llvm-project/releases/tag/llvmorg-10.0.0-rc2

Pre-built binaries will be added as they become available.

Please file bug reports for any issues you find as blockers of

Release testers: please run the test script, share your results, and
upload binaries.

I'm hoping we can now start tying up the loose ends, fixing the
blocking bugs, and getting the branch ready for shipping as a stable
release soon.

Thanks,
Hans

2

Uploaded for Xubuntu/Ubuntu 19.10.

sha256sum clang+llvm-10.0.0-rc2-x86_64-pc-linux-gnu.tar.xz
620f3531a337cab102e97dc4cd6d14fd8aaa015040cc160f8c8992ee25d38a79

Obtained many fail lines with the following text.
clang frontend command failed due to signal (use -v to see invocation)

Did not report the ParallelSTL fail lines and other apparently not relevant lines.

Testing Time: 674.25s
Expected Passes : 67683
Expected Failures : 270
Unsupported Tests : 1949
[100%] Built target check-all

make[2]: Target ‘CMakeFiles/check.dir/all’ not remade because of errors.
make[1]: *** [CMakeFiles/Makefile2:2483: CMakeFiles/check.dir/rule] Error 2
make[1]: Target ‘check’ not remade because of errors.
make: *** [Makefile:165: check] Error 2
[Release+Asserts Phase3] test suite failed

Packaging the release as clang+llvm-10.0.0-rc2-x86_64-pc-linux-gnu.tar.xz

Testing Finished

Logs: /home/nnelson/Documents/llvm-project/llvm/utils/release/rc2/logs

Errors:

[Release Phase3] test suite failed
[Release+Asserts Phase3] test suite failed

Neil Nelson

3

Uploaded ubuntu 18 binaries.

$ cat clang+llvm-10.0.0-rc2-x86_64-linux-gnu-ubuntu-18.04.tar.xz.sha256
8ca2cd0e0ba2243c095134373b46ccad822192b0495ce13eb0af33e12f9d17e1 clang+llvm-10.0.0-rc2-x86_64-linux-gnu-ubuntu-18.04.tar.xz

4

Windows binaries are ready. They were built with the attached batch file.

$ sha256sum LLVM-10.0.0-rc2-win*.exe
899b5ce0844809642e01fdc5e14c91a19f372ca588033ac506e56d1d6dac0bfe
LLVM-10.0.0-rc2-win32.exe
1e4d19759e6adb27e474f1624e19101f945cd97e0d4603d18a5f9a3f6b597b77
LLVM-10.0.0-rc2-win64.exe

build_llvm_1000-rc2.bat|attachment (4.58 KB)

5

Uploaded for Xubuntu/Ubuntu 19.10.

sha256sum clang+llvm-10.0.0-rc2-x86_64-pc-linux-gnu.tar.xz
620f3531a337cab102e97dc4cd6d14fd8aaa015040cc160f8c8992ee25d38a79

Obtained many fail lines with the following text.
clang frontend command failed due to signal (use -v to see invocation)

That doesn't sound so good. Can you share the big log file from
testing (logs/teting.test-*.log) if you still have it around?

6

Thanks! Added this and the other binaries I've gotten so far to the
release page and github.

7

This time I've switched to python3.8, and I've noticed that libc++ & co.
tests fail with it. This is already fixed in master, and the fix is
trivial, so I've requested backport:
https://bugs.llvm.org/show_bug.cgi?id=44905

No regressions found on my end.

8

For this rc, I used three patches, which are attached.

Main results on amd64-freebsd11:

  Expected Passes : 67927
  Expected Failures : 266
  Unsupported Tests : 4654
  Unexpected Passes : 5
  Unexpected Failures: 539
  Individual Timeouts: 19

Main results on i386-freebsd11:

  Expected Passes : 64981
  Passes With Retry : 1
  Expected Failures : 249
  Unsupported Tests : 3083
  Unresolved Tests : 1
  Unexpected Passes : 5
  Unexpected Failures: 231
  Individual Timeouts: 9

Uploaded:
SHA256 (clang+llvm-10.0.0-rc2-amd64-unknown-freebsd11.tar.xz) = 5d2be46c05f1db55391caec2abfea2558121eaed41ab586281555a0f337e3e1b
SHA256 (clang+llvm-10.0.0-rc2-i386-unknown-freebsd11.tar.xz) = 203e45a8ded937a6fc2859475329bad1e336676bfc3ad693f96ad52d3693b2ed

On both amd64 and i386, the test-suite build still segfaults (and even
results in one clang-10 process hanging at 100% CPU), which is tracked
in https://bugs.llvm.org/show_bug.cgi?id=44763.

-Dimitry

fix-clang-1.diff (447 Bytes)

fix-compiler-rt-1.diff (890 Bytes)

fix-test-suite-1.diff (552 Bytes)

9

Hello,

I finished testing llvm-10.0.0 RC2 on Power PC 64bit Little Endian Red Hat 7.4 machine and have uploaded the binary from IBM.

There were no failures. The sha1 file is attached.

Thanks,

Anil Mahmud.

clang+llvm-10.0.0-rc2-powerpc64le-linux-rhel-7.4.sha1 (98 Bytes)

10

Hi,

Results for 32 bits ARM are :
  Expected Passes : 63679
  Expected Failures : 261
  Unsupported Tests : 2739
  Unexpected Failures: 23

Failures are still cfi (PR44157) and Asan (PR44158) ones.

Results for AArch64 are:
  Expected Passes : 65349
  Expected Failures : 254
  Unsupported Tests : 3385
  Unexpected Failures: 2

The two failures are:
    AddressSanitizer-aarch64-linux :: TestCases/Posix/waitid.cpp
    AddressSanitizer-aarch64-linux-dynamic :: TestCases/Posix/waitid.cpp

with output error:

/usr/bin/ld: warning:
rc2/Phase3/Release/llvmCore-10.0.0-rc2.obj/lib/clang/10.0.0/lib/linux/libclang_rt.asan_cxx-aarch64.a(ubsan_type_hash_win.cpp.o):
unsupported GNU_PROPERTY_TYPE (5) type: 0xc0000000
rc2/llvm-project/compiler-rt/test/asan/TestCases/Posix/waitid.cpp:22:12:
error: CHECK: expected string not found in input
// CHECK: {{in .*waitid}}
           ^
<stdin>:4:1: note: scanning from here
AddressSanitizer:DEADLYSIGNAL
^
<stdin>:4:2: note: possible intended match here
AddressSanitizer:DEADLYSIGNAL
^

* Uploaded sha256sum binaries are:

e4a1c7dc1fd839ad1f40bb158852773ae5498f86eae0b9c3a734ff5f9684a905
clang+llvm-10.0.0-rc2-aarch64-linux-gnu.tar.xz
9f9107b393f3cc0c904fab3faf1acd187c128a4e23f0a95dc72617817c7e3eba
clang+llvm-10.0.0-rc2-armv7a-linux-gnueabihf.tar.xz

Thanks,
Yvan

11

Thanks! The Python fix has been merged now.

12

Here is the macOS build:

sha256 sum: b845b39ac60d4000bd3694f2de731ba15fb42e10a87ab6e6613ba95da982c982

Seems like a lot of tests where fixed between rc1 and rc2:

FAIL: LLVM :: Bindings/Go/go.test (25500 of 67049)
FAIL: ThreadSanitizer-x86_64 :: Darwin/norace-objcxx-run-time.mm
(57221 of 67049)
FAIL: libunwind :: signal_frame.pass.cpp (64756 of 67049)

(Can we fix my sftp access?)

-- Tobias

13

Hello,

I finished testing llvm-10.0.0 RC2 on Power PC 64bit Little Endian Ubuntu 16.04 machine and have uploaded the binary from IBM.

The sha1 file is attached.

When using ninja (-use-ninja) with test-release.sh, there were no failures. But when I did not use ninja the following failures were encountered both

in Release & Release+Asserts Flavors.

make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/alti.expandfft’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/alti.isamax’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/ldl’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/lvsr’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/mult-even-odd’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/pack’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/splat’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/st’, needed by ‘CMakeFiles/check’.
make[3]: *** No rule to make target ‘SingleSource/UnitTests/Vector/Altivec/ste’, needed by ‘CMakeFiles/check’.
make[3]: Target ‘CMakeFiles/check.dir/build’ not remade because of errors.
CMakeFiles/Makefile2:610: recipe for target ‘CMakeFiles/check.dir/all’ failed
make[2]: *** [CMakeFiles/check.dir/all] Error 2
CMakeFiles/Makefile2:617: recipe for target ‘CMakeFiles/check.dir/rule’ failed
make[1]: *** [CMakeFiles/check.dir/rule] Error 2
make[1]: Target ‘check’ not remade because of errors.
Makefile:164: recipe for target ‘check’ failed
make: *** [check] Error 2

  • deferred_error 3 ReleaseAndAsserts ‘test suite failed’
  • Phase=3
  • Flavor=ReleaseAndAsserts
  • Msg=‘test suite failed’
  • echo ‘[ReleaseAndAsserts Phase3] test suite failed’
  • tee -a /home/anil/llvmUpload-2020-02-18_16-32-44/rc2/logs/deferred_errors.log
    [ReleaseAndAsserts Phase3] test suite failed

Comparing Phase 2 and Phase 3 files

Packaging the release as clang+llvm-10.0.0-rc2-powerpc64le-linux-ubuntu-16.04.tar.xz

Testing Finished

Logs: /home/anil/llvmUpload-2020-02-18_16-32-44/rc2/logs

Errors:

[Release Phase3] test suite failed
[ReleaseAndAsserts Phase3] test suite failed

Thanks,

Anil Mahmud.

clang+llvm-10.0.0-rc2-powerpc64le-linux-ubuntu-16.04.sha1 (102 Bytes)

14

Hello,

The S390X bug reported in is fixed, thanks!

I am facing this issue (

Some other Debian/Ubuntu specific issues on some archs but no big deal!

Cheers,
Sylvestre

15

Hi,

due to the filing of http://llvm.org/PR45001 I was made be aware that
we could face a flood of emails about Polly not working anymore. We
could avoid that by merging https://reviews.llvm.org/D72372.
Otherwise, as mentioned in http://llvm.org/PR45001, we'd have to
rewrite the documentation about how to build Polly.

I know this is not ideal two days before the planned release. @zmodem
What approach do you prefer? So far I've added 45001 to the release
blockers.

Michael

16

That sounds like an invasive change.

Can we revert the change that broke the Polly build instead?

17

That sounds like an invasive change.

It changes a default value to what it was in llvm-9. I'd consider it
relatively risk-free.

Can we revert the change that broke the Polly build instead?

That'd be https://reviews.llvm.org/rG24ab9b537e61b3fe5e6a1019492ff6530d82a3ee
which is an invasive change. Serge Guelton already wrote some some
fixes on top of that, others such as https://bugs.llvm.org/PR44870 are
still unfixed.

If you think this is easier, then I'd say go ahead.

Michael

18

Hi,

I added the change to Polly's release note in
https://github.com/llvm/llvm-project/commit/002af0119286297dbd76b08a4a6cc4b6b87d3f26
to be sure that we have some documentation, in case you decide to tag
the release as-is.

In case you decide to either revert
24ab9b537e61b3fe5e6a1019492ff6530d82a3ee or cherry-pick D72372 (I'd
prefer either over keeping as-is), could you undo the release note
change as well?

Michael

19

I've gone ahead and cherry-picked D72372 and reverted the release note.

Thanks,
Hans

20

Thank you.

Michael