Re: [AMBER-Developers] false test failures

From: Jason Swails <jason.swails.gmail.com>
Date: Sun, 11 Apr 2010 01:03:56 -0400

Hello,

The old Makefile_at did indeed check for AMBERHOME as Scott pointed
out, but since the scripts now invoke make with -k (to make errors
non-fatal, so is_amberhome_defined is now obsolete), the tests are
plowing through anyway, so AMBERHOME existence should probably be
tested for in the scripts themselves (indeed for the AmberTools test
scripts it is). I've attached a quick patch (applied from AMBERHOME)
that does this. It could also be taken a step further. The patch
uses "dirname" as it's used in the configure script to determine what
AMBERHOME should be. Given the confusion regarding AMBERHOME and the
previous release of AmberTools (1.3 in October), perhaps we ought to
require that AMBERHOME is equal to that string and emit an error
message if not?

To include it? Not to? Not a question for me to answer, but here is
the patch nonetheless (I'll change it to require equivalence to the
determined directory if so desired).

All the best,
Jason

On Sun, Apr 11, 2010 at 12:05 AM, Sally Pias <sallypias.gmail.com> wrote:
> Just wanted to confirm that defining AMBERHOME eliminated the problem
> of false serial test failures (with empty *.dif files).
>
> Have the test scripts been updated to check whether AMBERHOME is
> defined and to print an error if not?
>
>
> On Thu, Apr 8, 2010 at 10:49 PM, Scott Brozell <sbrozell.rci.rutgers.edu> wrote:
>> Hi,
>>
>> We were; transplant target is_amberhome_defined from test/Makefile_at
>>
>> scott
>>
>> On Thu, Apr 08, 2010 at 12:02:24PM -0700, Ross Walker wrote:
>>>
>>> I see this behavior if I forget to set AMBERHOME or have it set to an
>>> incorrect directory.
>>>
>>> We should probably be checking for this in the test scripts.
>>>
>>> > On Thu, Apr 08, 2010, Sally Pias wrote:
>>> >
>>> > > I have successfully compiled the March 31 release candidate using
>>> > > Intel 11.1 compilers.  In running the serial test suite, I have
>>> > > repeatedly gotten the following results:
>>> > >
>>> > > Finished serial test suite for Amber 11 at Thu Apr  8 02:52:38 EDT
>>> > 2010.
>>> > >
>>> > > 315 file comparisons passed
>>> > > 95 file comparisons failed
>>> > > 2 tests experienced errors
>>> > >
>>> > > Of the 95 failed file comparisons, 93 are "false," in that the *.dif
>>> > > files are empty.  Any idea what is going on here?
>>> >
>>> > Nope...I don't see this behavior, nor to others...see the wiki page.
>>
>> _______________________________________________
>> AMBER-Developers mailing list
>> AMBER-Developers.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>



-- 
---------------------------------------
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
352-392-4032



_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers


Received on Sat Apr 10 2010 - 22:30:03 PDT
Custom Search