yep

On Wed, Sep 9, 2009 at 6:35 PM, Ethan Mallove <ethan.mall...@sun.com> wrote:

> On Wed, Sep/09/2009 10:30:44AM, Mike Dubman wrote:
> >    Hey Eytan,
> >
> >    It seems argv is participating in the following scenarios:
> >
> >    1. argv should be defined in mtt.ini for every single [Test Run]
> section
> >    2. Currently, &test_argv() is returing un-evaluated argv`s value
> >    3. &test_argv() is usually part of "exec=" parameter line of [MPI
> >    Details], which is evaluated for very test invocation:
> >
> >    mpiexec @options@ -n &test_np() &test_executable() &test_argv()
> >
> >    According to analysis above, if argv contains funclets or variables,
> they
> >    will get expanded during "exec" line evaluation.
>
> Okay, so we delay the evaluation of "argv" to when "exec" is
> evaluated. The error case is:
>
>  argv = &test_np()
>
> Before fix:
>
>  argv is undefined
>
> After fix:
>
>  argv is the value of &test_np()
>
> -Ethan
>
> >
> >    regards
> >
> >    Mike
> >
> >    On Tue, Sep 8, 2009 at 9:10 PM, Ethan Mallove <ethan.mall...@sun.com>
> >    wrote:
> >
> >      Mike,
> >
> >      What if argv contains a funclet, e.g.,
> >
> >      *argv = &foo()
> >
> >      Won't this change prevent it from getting expanded?
> >
> >      -Ethan
> >
> >      On Tue, Sep/08/2009 09:43:37AM, mi...@osl.iu.edu wrote:
> >      > Author: miked
> >      > Date: 2009-09-08 09:43:37 EDT (Tue, 08 Sep 2009)
> >      > New Revision: 1314
> >      > URL: https://svn.open-mpi.org/trac/mtt/changeset/1314
> >      >
> >      > Log:
> >      > fix:
> >      >
> >      > &test_np() can return incorrect value if used inside argv, here is
> a
> >      scenario:
> >      >
> >      > This behavior can be explained in next words as evaluation
> &np_test()
> >      > returns uninitialized $MTT::Test::Run::test_np that is initialized
> >      later in _run_one_np function.
> >      >
> >      > As a result using
> >      > $MTT::Test::Run::test_argv = $run->{argv};
> >      > allows to avoid damaging $MTT::Test::Run::test_argv *variable on
> >      current step but evaluation of &test_np() is done with whole
> >      command_line.
> >      >
> >      >
> >      > Text files modified:
> >      > * *trunk/lib/MTT/Test/RunEngine.pm | * * 2 +-
> >      > * *1 files changed, 1 insertions(+), 1 deletions(-)
> >      >
> >      > Modified: trunk/lib/MTT/Test/RunEngine.pm
> >      >
> >
>  
> ==============================================================================
> >      > --- trunk/lib/MTT/Test/RunEngine.pm * (original)
> >      > +++ trunk/lib/MTT/Test/RunEngine.pm * 2009-09-08 09:43:37 EDT
> (Tue, 08
> >      Sep 2009)
> >      > @@ -191,7 +191,7 @@
> >      > * * * * *$MTT::Test::Run::test_executable_abspath = $test_exe_abs;
> >      > * * * * *$MTT::Test::Run::test_executable_basename =
> >      $test_exe_basename;
> >      >
> >      > - * * * *$MTT::Test::Run::test_argv =
> >      MTT::Values::EvaluateString($run->{argv}, $ini,
> $test_run_full_name);
> >      > + * * * *$MTT::Test::Run::test_argv = $run->{argv};
> >      > * * * * *my $all_np = MTT::Values::EvaluateString($run->{np},
> $ini,
> >      $test_run_full_name);
> >      >
> >      > * * * * *my $save_run_mpi_details = $MTT::Test::Run::mpi_details;
> >      > _______________________________________________
> >      > mtt-svn mailing list
> >      > mtt-...@open-mpi.org
> >      > http://www.open-mpi.org/mailman/listinfo.cgi/mtt-svn
> >      _______________________________________________
> >      mtt-devel mailing list
> >      mtt-de...@open-mpi.org
> >      http://www.open-mpi.org/mailman/listinfo.cgi/mtt-devel
> >
> > References
> >
> >    Visible links
> >    . mailto:ethan.mall...@sun.com
> >    . mailto:mi...@osl.iu.edu
> >    . https://svn.open-mpi.org/trac/mtt/changeset/1314
> >    . mailto:mtt-...@open-mpi.org
> >    . http://www.open-mpi.org/mailman/listinfo.cgi/mtt-svn
> >    . mailto:mtt-de...@open-mpi.org
>  >    . http://www.open-mpi.org/mailman/listinfo.cgi/mtt-devel
>
> > _______________________________________________
> > mtt-devel mailing list
> > mtt-de...@open-mpi.org
> > http://www.open-mpi.org/mailman/listinfo.cgi/mtt-devel
>
> _______________________________________________
> mtt-devel mailing list
> mtt-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/mtt-devel
>

Reply via email to