[ovs-dev] openvswitch 2.0.0 build breaks on PPC

[Flavio Leitner]

On Wed, Oct 30, 2013 at 11:18:09AM -0700, Ben Pfaff wrote:
> On Wed, Oct 30, 2013 at 04:15:05PM -0200, Flavio Leitner wrote:
> > On Wed, Oct 30, 2013 at 10:28:31AM -0700, Ben Pfaff wrote:
> > > On Wed, Oct 30, 2013 at 10:33:31AM -0200, Flavio Leitner wrote:
> > > > On Tue, Oct 29, 2013 at 11:13:17AM -0700, Ben Pfaff wrote:
> > > > > On Tue, Oct 29, 2013 at 03:15:58PM -0200, Flavio Leitner wrote:
> > > > > > On Tue, Oct 29, 2013 at 08:37:58AM -0700, Ben Pfaff wrote:
> > > > > > > On Tue, Oct 29, 2013 at 10:51:02AM -0200, Flavio Leitner wrote:
> > > > > > > > While updating the packages to 2.0.0 I found that the build fails on
> > > > > > > > PPC.
> > > > > > > > 
> > > > > > > > I am not an expert on PPC arch, but I think it doesn't support atomic
> > > > > > > > operations of 8 bytes (64bit) on 32bit system, then there is no GCC
> > > > > > > > built-in[1] for those cases which breaks the build.
> > > > > > > 
> > > > > > > I only see build failures in the test-atomic program.  That's a good
> > > > > > > sign that we only use these 8-byte atomic operations in the test
> > > > > > > program, not in real code.  Are there other build failures?  (Does "make
> > > > > > > vswitchd/ovs-vswitchd" succeed?)
> > > > > > 
> > > > > > I've applied the patch below commenting out the problematic ones
> > > > > > in test-atomic to work around and I was able to build the rpms.
> > > > > > So, yeah, that's the only issue as far as I can see.
> > > > > 
> > > > > Did you run the OVS testsuite?  Any failures?
> > > > 
> > > > Yes, see below and the attached file testsuite.log.
> > > 
> > > In that case, there's significantly more going on than just 8-byte
> > > atomic operations in tests.  If you have time to dig into the
> > > problems, please go ahead, otherwise I will look next week or the week
> > > after.
> >  
> > You're saying that the other failures in the testsuite.log are
> > somehow related to the 8-byte atomic operations?
> 
> No, I'm saying that the other failures are unrelated to those and so
> they need to be investigated.  Removing the 8-byte atomic operations
> from the testsuite, by themselves, will not help
> 
> > I am seeing some tests failing on x86_64 as well, 34/1328.
> > They are 306~313, 1136, 1138~1162.
> 
> OK, could you send the testsuite.log for those?  I'm not aware of any
> test failures on x86_64.

Hm, that was my fault, sorry. I was just running tests/testsuite so
all the x86_64 failures were about not having the certificate files.
Of course, running 'make check' shows no errors.

Then I redid the ppc test. It's a lot better now.

ERROR: All 1328 tests were run,
3 failed unexpectedly.
## -------------------------- ##
## testsuite.log was created. ##
## -------------------------- ##

Please send `tests/testsuite.log' and all information you think might
help:

   To: <ovs-bugs at openvswitch.org>
   Subject: [openvswitch 2.0.0] testsuite: 2 6 13 failed

 NUM: FILE-NAME:LINE     TEST-GROUP-NAME
      KEYWORDS

   2: bfd.at:79          bfd - Verify tunnel down detection
   6: bfd.at:252         bfd - bfd decay
  13: lacp.at:121        lacp - negotiation

testsuite.log attached.

fbl
 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ppc-testsuite.log.gz
Type: application/gzip
Size: 39471 bytes
Desc: not available
URL: <http://mail.openvswitch.org/pipermail/ovs-dev/attachments/20131030/db1517ec/attachment-0005.bin>