octave-bug-tracker
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Octave-bug-tracker] [bug #57585] [octave-forge] (image) deprecation war


From: Hartmut
Subject: [Octave-bug-tracker] [bug #57585] [octave-forge] (image) deprecation warnings during installation (compilation) of image package
Date: Fri, 24 Jan 2020 11:01:36 -0500 (EST)
User-agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:72.0) Gecko/20100101 Firefox/72.0

Follow-up Comment #14, bug #57585 (project octave):

I have finished testing the dev-image package (as of yesterday evening,
including the proposed one-line patch of comment#13) under Window 10:

* Installing the packages works fine, on tested Octave versions that the image
package claims to support (i.e. Octave 5.1.90, 4.4.1 and 4.2.2)

* The image package's test suite runs without any errors on Octave 5.1.90 and
4.4.1.

* There is one single FAIL in the image test suite, in fftconv2:


>>>>> processing
C:\Octave\OCTAVE~1.2\share\octave\packages\image-2.11.0+\fftconv2.m
***** test
 a = repmat (1:10, 5);
 b = repmat (10:-1:3, 7);
 assert (fftconv2 (a, b), conv2 (a, b), 1.3e4*eps)
 assert (fftconv2 (b, a), conv2 (b, a), 1.3e4*eps)
 assert (fftconv2 (a, b, "full"), conv2 (a, b, "full"), 1.3e4*eps)
 assert (fftconv2 (b, a, "full"), conv2 (b, a, "full"), 1.3e4*eps)
 assert (fftconv2 (a, b, "same"), conv2 (a, b, "same"), 1e4*eps)
 assert (fftconv2 (b, a, "same"), conv2 (b, a, "same"), 1e4*eps)
 assert (isempty (fftconv2 (a, b, "valid")));
 assert (fftconv2 (b, a, "valid"),  conv2 (b, a, "valid"), 1e4*eps)
!!!!! test failed
ASSERT errors for:  assert (fftconv2 (a, b),conv2 (a, b),1.3e4 * eps)

  Location  |  Observed  |  Expected  |  Reason
   (6,58)    9100+5.4622766e-013i     9100       Abs err 3.6788e-012 exceeds
tol 2.8866e-012
   (7,58)    9100+5.701925e-013i     9100       Abs err 3.6824e-012 exceeds
tol 2.8866e-012


Those tolerances were already loosened a bit for Windows, see bug #45108.
(from 1e4*eps to 1.3e4*eps).

My suggestion would be, to further loosen the tolerances, for example from
1.3e4*eps to 1.8e4*eps or maybe 2e4*eps (1.6e4 is not enough). Any thoughts or
objections to this? 

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?57585>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/




reply via email to

[Prev in Thread] Current Thread [Next in Thread]