[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-plotutils] ploututils testsuite failure on spline (i386)
From: |
Gianfranco Costamagna |
Subject: |
Re: [Bug-plotutils] ploututils testsuite failure on spline (i386) |
Date: |
Thu, 28 Apr 2016 10:55:55 +0000 (UTC) |
Hi Robert, thanks for the quick and complete answer!
by looking at the numbers I was wondering about just disabling the testsuite on
i686
(I don't see any pratical difference between -9.5518e-17 and 0 in the real
world :) )
I uploaded a fix in Ubuntu with the testsuite disabled on that arch, thanks for
confirming
and the quick "fix"
cheers,
G.
Il Giovedì 28 Aprile 2016 1:10, Robert S. Maier <address@hidden> ha scritto:
On 04/26/2016 07:27 AM, Gianfranco <address@hidden> wrote:
> Hi, I'm attaching the spline.test directory and the build log on
Ubuntu yakkety i386
> (the default should be i686 now) thanks, Gianfranco
Gianfranco, thanks for the bug report on `spline' in plotutils. It's not
at all serious, fortunately.
What's happening is this: the `make test' procedure for plotutils
executes the `spline.test' script in the test directory, which runs the
command
echo 0 1 0.5 0.75 -0.25 0 | ../spline/spline -a -T 1 -n 120 --precision
5 -p >spline.out
The output file, spline.out, is then compared against the expected
output, which is in the file spline.xout. They don't exactly match on
your system, which is why the test failed.
If you do a `diff spline.xout spline.out' to compare the two files,
you'll see this:
1c1
< 0 0
---
> 0 I
121c121
< 5 0
---
> 5 -9.5518e-17
The difference between -9.5518e-17 and zero (which are y-values
generated by `spline', at the endpoints of an interpolatory range) is
not a major one!
However, this ought to be fixed in the next release. (This numerical
bug appears on some systems but not on others, for unknown reasons; so
the easiest way to deal with it would be to modify the test script
`spline.test', to ignore the first and last lines when comparing
spline.out with spline.xout.)
Best, Rob