[llvm] r227414 - [LPM] Try again to appease powerpc64 in its self host. I've been unable

[Hal Finkel]

----- Original Message -----
> From: "Chandler Carruth" <chandlerc at gmail.com>
> To: "Hal Finkel" <hfinkel at anl.gov>
> Cc: "Chandler Carruth" <chandlerc at gmail.com>, "Commit Messages and Patches for LLVM" <llvm-commits at cs.uiuc.edu>
> Sent: Wednesday, January 28, 2015 9:58:29 PM
> Subject: Re: [llvm] r227414 - [LPM] Try again to appease powerpc64 in its self host. I've been unable
> 
> On Wed, Jan 28, 2015 at 7:19 PM, Hal Finkel < hfinkel at anl.gov >
> wrote:
> 
> 
> ----- Original Message -----
> > From: "Chandler Carruth" < chandlerc at gmail.com >
> > To: llvm-commits at cs.uiuc.edu
> > Sent: Wednesday, January 28, 2015 8:34:17 PM
> > Subject: [llvm] r227414 - [LPM] Try again to appease powerpc64 in
> > its self host. I've been unable
> > 
> > Author: chandlerc
> > Date: Wed Jan 28 20:34:17 2015
> > New Revision: 227414
> > 
> > URL: http://llvm.org/viewvc/llvm-project?rev=227414&view=rev
> > Log:
> > [LPM] Try again to appease powerpc64 in its self host. I've been
> > unable
> > to get a powerpc64 host so that I can reproduce and test this, but
> > it
> > only impacts that platform so trying the only other realistic
> > option.
> > 
> > According to Ulrich, who debugged this initially, initial-exec is
> > likely
> > to be sufficient for our needs and not subject to this bug. Will
> > watch
> > the build bots to see.
> > 
> > If this doesn't work, I'll be forced to cut a really ugly
> > pthread-based
> > approach into the primary user (our stack trace printing) as that
> > user
> > cannot use the ThreadLocal implementation due to lifetime issues.
> 
> What about if we added -Wl,--no-tls-optimize
> -Wl,--no-tls-get-addr-optimize to our self-hosting build flags for
> now?
> 
> -Hal
> It happens that I managed to get a self host going on a PPC machine,
> and initial-exec fixed the issue for me. But not the build bot! Oh
> no, that one is still broken.
> http://lab.llvm.org:8011/builders/clang-ppc64-elf-linux2/builds/17257/steps/compile.llvm.stage2/logs/stdio
> 
> 
> 
> I am at a loss, and not really competent to propose a fix as even
> with the only ppc machine I have, I no longer can reproduce this.

I was able to reproduce the problem with r227415 on my PPC64 P7 box, and setting:
  CMAKE_EXE_LINKER_FLAGS:STRING=-Wl,--no-tls-get-addr-optimize -Wl,--no-tls-optimize
  CMAKE_MODULE_LINKER_FLAGS:STRING=-Wl,--no-tls-get-addr-optimize -Wl,--no-tls-optimize
made the problem go away. Given that the bug is actually in the linker, I think this is really the proper solution. Can we add these to our build flags for PPC64 for now?

 -Hal

-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory