6.0.1-rc2 has been tagged

Project Infrastructure Release Testers
1

Hi,

The 6.0.1-rc2 release has been tagged. Testers may begin testing and
reporting results.

-Tom

2

Windows looks good and binaries are ready:

$ sha1sum LLVM-6.0.1-rc2*
e09cde078ab3a90c2db5e1a7ece9b053b873ed5f LLVM-6.0.1-rc2-win32.exe
dd483a84183c8d9a10ac81e8efce806d1a768bdc LLVM-6.0.1-rc2-win64.exe

They were built with the attached batch file.

Thanks,
Hans

build_llvm_601.bat|attachment (4.64 KB)

3

Built for FreeBSD 10, tested and uploaded:

SHA256 (clang+llvm-6.0.1-rc2-amd64-unknown-freebsd10.tar.xz) = b2858b1bf254f9aed977638229d90e3a8aeb245e9f12156bda5597cd82e6ae20
SHA256 (clang+llvm-6.0.1-rc2-i386-unknown-freebsd10.tar.xz) = 5d42eb7022a1974edcb74091b93a6c6eaa30cac12ea6f1f93324577b5e2a4261

4

Green on Debian on all supported archs.

Do you mind if I merge this patch:

in the 6.0 branch?

Thanks
S

5

Are these regressions?

-Tom

6

For i386, one failure (Bindings/Go/go.test) was fixed, and no new failures were introduced:

rc1> Failing Tests (247):
rc2> Failing Tests (246):
rc1> LLVM :: Bindings/Go/go.test
rc1> Expected Passes : 44251
rc2> Expected Passes : 44268
rc1> Unexpected Failures: 247
rc2> Unexpected Failures: 246

For amd64, one failure (Bindings/Go/go.test) was fixed, but a new failure (api/omp_get_wtime.c) was introduced:

rc1> LLVM :: Bindings/Go/go.test
rc2> libomp :: api/omp_get_wtime.c
rc1> Expected Passes : 45407
rc2> Expected Passes : 45423

It's not immediately clear why the omp_get_wtime test fails:

FAIL: libomp :: api/omp_get_wtime.c (47166 of 49069)
******************** TEST 'libomp :: api/omp_get_wtime.c' FAILED ********************
Script:

7

This test fails in rare cases when you heavily overload your system because it compares measured timings. Because of that the check allows quite some deviation that is usually enough. Maybe we should think about a better way to test this...

8

...

For amd64, one failure (Bindings/Go/go.test) was fixed, but a new
failure (api/omp_get_wtime.c) was introduced:
rc1> LLVM :: Bindings/Go/go.test
rc2> libomp :: api/omp_get_wtime.c
rc1> Expected Passes : 45407
rc2> Expected Passes : 45423
It's not immediately clear why the omp_get_wtime test fails:
FAIL: libomp :: api/omp_get_wtime.c (47166 of 49069)
******************** TEST 'libomp :: api/omp_get_wtime.c' FAILED

...

This test fails in rare cases when you heavily overload your system because it compares measured timings. Because of that the check allows quite some deviation that is usually enough. Maybe we should think about a better way to test this...

Indeed, I re-ran the whole build, and now this test passed. Let's write
it off as a fluke. :slight_smile:

-Dimitry

9

Hi,

The 6.0.1-rc2 release has been tagged. Testers may begin testing and
reporting results.

Green on Debian on all supported archs.

Do you mind if I merge this patch:
debian/patches/ubuntu-cosmic-support.patch · 6.0 · pkg-llvm / llvm-toolchain · GitLab

in the 6.0 branch?

How important is this patch? Does clang work on cosmic without it?

-Tom

10

Hi

MIPS looks ok. I suffered some specific build machine configuration issues but otherwise it looks ok. Binaries uploaded.

SHA256(clang+llvm-6.0.1-rc2-mipsel-linux-gnu.tar.xz)= 2573c0ccc42ecb7f5b36aa000245265f389ace1f174805d2ad46a2d33b9c0533
SHA256(clang+llvm-6.0.1-rc2-mips-linux-gnu.tar.xz)= f09f8a2dc7dab15a399f0b5bf48a5f21a82c6c896a2e5f7c086112b3ee177e5d
SHA256(clang+llvm-6.0.1-rc2-x86_64-linux-gnu-debian8.tar.xz)= a83e05f3b609afc569ef0f0ec3e354d152db39a1c1cc142c5b36bb7b848d8f34

Thanks,
Simon

11

Missed release-testers.

Simon

12

I don't think it is a big deal but as the patch is trivial, would not
hurt for maintenance.

Sylvestre

13

Hi,

The 6.0.1-rc2 release has been tagged. Testers may begin testing and
reporting results.

Green on Debian on all supported archs.

Do you mind if I merge this patch:
debian/patches/ubuntu-cosmic-support.patch · 6.0 · pkg-llvm / llvm-toolchain · GitLab

in the 6.0 branch?

How important is this patch? Does clang work on cosmic without it?

I don't think it is a big deal but as the patch is trivial, would not
hurt for maintenance.

I think it's too late for this now. I would like to get this release
out the door without having to do another -rc candidate if we can
avoid it.

-Tom