3.5.1 Testing Phase Begins

Hi,

3.5.1-rc1 has been tagged, so it is time to begin testing. We can
always use more testers, so if you would like to volunteer, let me
know.

For those that are new to testing, please review the documentation
for how validate a new release:
http://llvm.org/docs/ReleaseProcess.html

Remember that when we check for regressions we want to compare the
test results of 3.5.1-rc1 with 3.5.0-final.

There is one undocumented step, which is how to package build/package
the binaries so they can be uploaded to the website. I think there
may be a script for this, but I'm not sure. It would be great if
one of the veteran testers could add this information to the release
documentation, or at the very least explain to me what to do so
I can update the docs.

If you have any questions about testing, don't hesitate to ask.

Thanks,
Tom

test-release.sh outputs the compressed archives and we just upload them to ftp. I think you just have to move them somewhere and link to them from the webpage.

test-release.sh outputs the compressed archives and we just upload them to
ftp. I think you just have to move them somewhere and link to them from the
webpage.

What directory does test-release.sh put the compressed archives?

-Tom

Wherever you invoke it from.

Testing ARMv7 and AArch64. Will upload when it finishes.

cheers,
--renato

Sorry for coming late to the party; my Windows box decided to die
earlier this week.

After re-installing everything, lld refused to build on the 3.5
branch. I'm not sure how it ever worked before, but I need to merge
r220117 to make it work.

I also got two test failures. I haven't checked yet if they also occur
in 3.5, but I'm not sure they matter on Windows:

LLVM: CodeGen/Mips/cconv/arguments-struct.ll
Clang: Modules/compiler_builtins.m

I've uploaded it to the sftp:
LLVM-3.5.1-rc1-win32.exe
sha1: 3c3892daa5119e60ee7b0917c4c1ac8251a949c6

- Hans

You’re not the only one Hans, my VMs have started crashing my host OS :slight_smile:

Fedora and OpenSUSE look good, binaries uploaded.

Same here... ARMv7 is already up and tested, AArch64 is proving a bit
more complicated due to old libstdc++ on my box. I'm working on it.

cheers,
--renato

I'm seeing one error on AArch64's test-suite:

SingleSource/UnitTests/Vector/AArch64/aarch64_neon_intrinsics.c:154979:16:
warning: implicit declaration of function 'vmovq_n_p64' is invalid in
C99 [-Wimplicit-function-declaration]
      result = vmovq_n_p64(arg1);
               ^
SingleSource/UnitTests/Vector/AArch64/aarch64_neon_intrinsics.c:154979:14:
error: assigning to 'poly64x2_t' (vector of 2 'poly64_t' values) from
incompatible type 'int'
      result = vmovq_n_p64(arg1);
             ^ ~~~~~~~~~~~~~~~~~
1 warning and 1 error generated.

Tim, any ideas?

cheers,
--renato

Tim, any ideas?

Isn't that the test added by Kristof recently (based on the emperor
suite)? Are you using the release_35 branch of test-suite (I didn't
even know such a thing existed until 30s ago, good thinking whoever
did that!).

Cheers.

Tim.

That’s a good point, I think I’m using trunk test-suite… I’ll check.

Cheers,
Renato

Everything else seems fine, so I uploaded the AArch64 tar ball, too.

cheers,
--renato

Hi everyone,

I have a failing test on OS X 10.10:

FAIL: LLVM :: CodeGen/Mips/cconv/arguments-struct.ll (9584 of 18783)
******************** TEST 'LLVM :: CodeGen/Mips/cconv/arguments-struct.ll' FAILED ********************
Script:

LLVM: CodeGen/Mips/cconv/arguments-struct.ll

Do you have the verbose lit output for this test?

I ran it again in verbose mode, but there were no additional outputs.

Cheers,
Sebastian

From: llvmdev-bounces@cs.uiuc.edu [mailto:llvmdev-bounces@cs.uiuc.edu]
On Behalf Of Sebastian Dreßler
Sent: 15 December 2014 14:08
Cc: Ben Pope; Nikola Smiljanic; llvmdev
Subject: Re: [LLVMdev] 3.5.1 Testing Phase Begins

>> LLVM: CodeGen/Mips/cconv/arguments-struct.ll
>
> Do you have the verbose lit output for this test?
>

I ran it again in verbose mode, but there were no additional outputs.

Cheers,
Sebastian

Just to double check: Is that with -v or -vv? The latter should print the output of the failing command.

It was -v but I ran again with -vv. However, there is no new information.

For sake of completeness, I copy the output again:

Command Output (stderr):

From: sebastian.dressler@gmail.com [sebastian.dressler@gmail.com]
Sent: 15 December 2014 18:41
To: Daniel Sanders
Cc: Ben Pope; Nikola Smiljanic; llvmdev
Subject: Re: [LLVMdev] 3.5.1 Testing Phase Begins

>
>> From: llvmdev-bounces@cs.uiuc.edu [mailto:llvmdev-bounces@cs.uiuc.edu]
>> On Behalf Of Sebastian Dreßler
>> Sent: 15 December 2014 14:08
>> Cc: Ben Pope; Nikola Smiljanic; llvmdev
>> Subject: Re: [LLVMdev] 3.5.1 Testing Phase Begins
>>
>>>> LLVM: CodeGen/Mips/cconv/arguments-struct.ll
>>>
>>> Do you have the verbose lit output for this test?
>>>
>>
>> I ran it again in verbose mode, but there were no additional outputs.
>>
>> Cheers,
>> Sebastian
>
> Just to double check: Is that with -v or -vv? The latter should print the output of the failing command.

It was -v but I ran again with -vv. However, there is no new information.

For sake of completeness, I copy the output again:

Command Output (stderr):
--
/Users/rock/llvm/3.5.1-testing/rc1/llvm.src/test/CodeGen/Mips/cconv/arguments-struct.ll:28:14: error: expected string not found in input
; SYM32-DAG: lui [[PTR_HI:\$[0-9]+]], %hi(bytes)
             ^
<stdin>:15:7: note: scanning from here
s_i8: # @s_i8
      ^
<stdin>:24:5: note: possible intended match here
addiu $2, $gp, %gp_rel(bytes)
    ^

Cheers,
Sebastian

That's given me enough to figure out the problem. %gp_rel is only used when the small data section is enabled and that's only supposed to be the case for bare-metal. However, the guard is actually a test for 'is not Linux' and this condition is of course true for windows.

It's not a new bug (it was in 3.5.0 too) but the testcase is new. Does the attached patch fix the problem for you?

arguments-struct.patch (2.79 KB)

From: sebastian.dressler@gmail.com [sebastian.dressler@gmail.com]
Sent: 15 December 2014 18:41
To: Daniel Sanders
Cc: Ben Pope; Nikola Smiljanic; llvmdev
Subject: Re: [LLVMdev] 3.5.1 Testing Phase Begins

For sake of completeness, I copy the output again:

Command Output (stderr): […]

That’s given me enough to figure out the problem. %gp_rel is only used when the small data section is enabled and that’s only supposed to be the case for bare-metal. However, the guard is actually a test for ‘is not Linux’ and this condition is of course true for windows.

Just out of curiosity: this information was also contained in my initial mail, did you probably overlook it?

It’s not a new bug (it was in 3.5.0 too) but the testcase is new. Does the attached patch fix the problem for you?<arguments-struct.patch>

Yes, it fixes it.

Cheers,
Sebastian

Hi,

uploaded binary to SFTP: clang+llvm-3.5.1-rc1-x86_64-apple-darwin.tar.xz

Except for the failing test mentioned before, there were no regressions.

Cheers,
Sebastian