compile errors

[Rob Gowin]

On Sat, Nov 19, 2016 at 6:18 PM, John Ralls <jralls at ceridwen.us> wrote:

>
> > On Nov 19, 2016, at 11:14 AM, Rob Gowin <robg at gowin.net> wrote:
> >
> > On Sat, Nov 19, 2016 at 12:26 PM, Alex Aycinena <alex.aycinena at gmail.com
> >
> > wrote:
> >
> >> On Sat, Nov 19, 2016 at 10:09 AM, John Ralls <jralls at ceridwen.us>
> wrote:
> >>
> >>>
> >>>> On Nov 19, 2016, at 9:34 AM, Alex Aycinena <alex.aycinena at gmail.com>
> >>> wrote:
> >>
> >>>>
> >>>> Don't know if you already knew that.
> >>>
> >>> Alex,
> >>>
> >>> Nope, didn't know that because it passes on Travis and all of my VMs
> >>> including Fedora-24, so ISTM you have a local problem.
> >>>
> >>> A clue is that the error about the error about xaccSplitEqualCheckBal
> >>> isn't issued until line 967, yet the test reports that it's failing at
> >> line
> >>> 948.
> >>>
> >>> Regards,
> >>> John Ralls
> >>>
> >>>
> >>> John,
> >>
> >> I think I got the error because I had done a git pull followed by a make
> >> and make check without redoing autogen, clearing out the build directory
> >> and rerunning configure. My mistake. When I did it that way, it all
> >> workied.
> >>
> >> Thanks,
> >>
> >> Alex
> >>
> >>
> >
> > [ Alex, I originally forgot to reply-all, so you'll see this twice. ]
> >
> > For what it's worth, I am able to reproduce this failure in test-engine
> on
> > Ubuntu 16.04.4, g++ 5.4.0. I started looking into why the test did not
> fail
> > in the CMake build environment. It seems to be an intermittent failure
> for
> > me.  It only happens when running under the gtester program, with both
> > MALLOC_CHECK_ and MALLOC_PERTURB_ set. After Alex's most recent comment,
> I
> > did a new clone, autogen.sh, configure, make and make check, and am able
> to
> > see the failure.
> >
> > The CMake test environment does not use gtester, nor does it set the
> > MALLOC* environment variables. I suppose it should.
> >
> >
> > I ran this script in src/test/engine to run test-engine 500 times and it
> > always fails at some point:
> >
> > #!/bin/bash
> >
> > for run in {1..500}; do
> >   MALLOC_CHECK_=2 MALLOC_PERTURB_=$((${RANDOM:-256} % 256)) gtester
> > --verbose test-engine
> >   if [ $? -ne 0 ]; then
> >       echo '** FAIL **'
> >       exit 1
> >   fi
> > done
> > echo '** PASS **'
> > exit 0
>
>
> Hmm. I ran your script 5 times (so 2500 total runs of test-engine) on my
> Fedora 24 VM and it passed every time.
>
> Does the machine/VM that you tested on have an installed GnuCash? If so,
> are you absolutely certain that the test build can't link the installed
> libraries?
>
> Regards,
> John Ralls
>

I'm using a machine (non-VM) that does not have GnuCash installed. However,
I built a clean Ubuntu 16.04 VM and I can't get it to fail in 2500 runs
either. Probably is something local to my box.

Rob