--testing-mode changes output of regression.sh

[Jason Stover-3]

I was fixing bug 23567 and ran make check. I found the results were
very different from those obtained by running the same syntax without
--testing-mode. Does anyone know why this should happen?

-Jason


_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev

[Jason Stover-3]

On Wed, Jun 18, 2008 at 06:45:43PM -0400, Jason Stover wrote:
> I was fixing bug 23567 and ran make check. I found the results were
> very different from those obtained by running the same syntax without
> --testing-mode. Does anyone know why this should happen?

http://math.gcsu.edu/~jhs/pspp.tar.bz2 if anyone wants to reproduce the
problem. Check the syntax in regression.sh with and without --testing-mode.



_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev

[John Darrington]

When you say the results are different, do you mean different values
are produced, or just formatted differently?

J'

On Wed, Jun 18, 2008 at 06:55:10PM -0400, Jason Stover wrote:
     On Wed, Jun 18, 2008 at 06:45:43PM -0400, Jason Stover wrote:
     > I was fixing bug 23567 and ran make check. I found the results were
     > very different from those obtained by running the same syntax without
     > --testing-mode. Does anyone know why this should happen?
     
     http://math.gcsu.edu/~jhs/pspp.tar.bz2 if anyone wants to reproduce the
     problem. Check the syntax in regression.sh with and without --testing-mode.

--
PGP Public key ID: 1024D/2DE827B3
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.




_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev

[Jason Stover-3]

On Thu, Jun 19, 2008 at 09:35:31AM +0800, John Darrington wrote:
> When you say the results are different, do you mean different values
> are produced, or just formatted differently?

Different numerical values coming out of linreg.c. I'll look into it
more today.

-Jason



_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev

[Jason Stover-3]

On Thu, Jun 19, 2008 at 09:35:31AM +0800, John Darrington wrote:
> When you say the results are different, do you mean different values
> are produced, or just formatted differently?

--testing-mode gives this gsl_matrix to reg_sweep:

1.9111902193650001 -504.13075111149146  1552.549464380658
0                   19.278766355808902 -1662.0047084369501
0                   0                   215.25642906983995


Without that option, reg_sweep gets this matrix:

1.9111902193650001  2.4083892281683541 -4.7878197279351395
0                   19.278766355808902 -61.698877095504059
0                   0                   215.25642906983995

The second one is correct for the syntax in regression.sh.

I figure that I broke a gsl_matrix_view in linreg.c while fixing bug 23567, and
that's what I'm looking for now. What I'm wondering is why such
an error would show up with --testing-mode.

-Jason


_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev

[Jason Stover-3]

On Thu, Jun 19, 2008 at 05:12:41PM -0400, Jason Stover wrote:
> --testing-mode gives this gsl_matrix to reg_sweep:...

Forget my previous message. Now it looks like the problem has to
do with pointers to struct variables. They seem to get mixed up
in testing mode. I'll send another message about it when I finally
get this straightened out.

-Jason



_______________________________________________
pspp-dev mailing list
pspp-dev@...
http://lists.gnu.org/mailman/listinfo/pspp-dev