llvm dev - May 2015 - [LLVMdev] 3.6.1 -rc1 has been tagged. Testing begins.

On 14 May 2015 at 16:49, Renato Golin <renato.golin at linaro.org>
wrote:
> MultiSource/Benchmarks/MiBench/consumer-typeset/consumer-typeset
This was a fluke. The test passed later on all subsequent runs.

> MultiSource/UnitTests/C++11/frame_layout/frame_layout
This one is a new test, and I believe it wasn't there in 3.6.0, so
it's ok to fail, since the patch hasn't been backported.

$ ./frame_layout | grep NOT
+----, int_local, alignment: 4096 ALIGNMENT NOT AS EXPECTED: 0x7fec948520
+----, double_local, alignment: 4096 ALIGNMENT NOT AS EXPECTED: 0x7fec948510
...

Kristof,

Is this a fair assessment? If so, AArch64 3.6.1 is green.

cheers,
--renato

Renato, your assessment is spot on - the frame_layout test was
added to verify a bunch of frame layout aspects and indeed
the AArch64 backend didn't support over-aligning objects
on the stack on the 3.6 branch. This has been implemented
on trunk.

Thanks,

Kristof
> -----Original Message-----
> From: Renato Golin [mailto:renato.golin at linaro.org]
> Sent: 14 May 2015 17:31
> To: Tom Stellard
> Cc: LLVM Dev; Clang Dev; Kristof Beyls
> Subject: Re: [LLVMdev] 3.6.1 -rc1 has been tagged. Testing begins.
> 
> On 14 May 2015 at 16:49, Renato Golin <renato.golin at linaro.org>
wrote:
> > MultiSource/Benchmarks/MiBench/consumer-typeset/consumer-typeset
> 
> This was a fluke. The test passed later on all subsequent runs.
> 
> 
> > MultiSource/UnitTests/C++11/frame_layout/frame_layout
> 
> This one is a new test, and I believe it wasn't there in 3.6.0, so
it's
> ok to fail, since the patch hasn't been backported.
> 
> $ ./frame_layout | grep NOT
> +----, int_local, alignment: 4096 ALIGNMENT NOT AS EXPECTED:
> +0x7fec948520 ----, double_local, alignment: 4096 ALIGNMENT NOT AS
> +EXPECTED: 0x7fec948510
> ...
> 
> Kristof,
> 
> Is this a fair assessment? If so, AArch64 3.6.1 is green.
> 
> cheers,
> --renato

On 14 May 2015 at 17:50, Kristof Beyls <kristof.beyls at arm.com>
wrote:
> Renato, your assessment is spot on - the frame_layout test was
> added to verify a bunch of frame layout aspects and indeed
> the AArch64 backend didn't support over-aligning objects
> on the stack on the 3.6 branch. This has been implemented
> on trunk.
Perfect, thanks!

The binary in the SFTP site is good to go, then.

cheers,
--renato