[Gretl-devel] Re: PRE-RELEASE: Testscripts fail
On Wed, Jul 12, 2023 at 2:39 AM Artur T. <atecon(a)posteo.de> wrote:
Hi,
Given Allin's announcement that a new version will be released soon, I
ran the test scripts.
1. For the following script the assertion failed on the latest git-version:
https://sourceforge.net/p/gretl/workspace/ci/master/tree/tests/test_scrip...
I am pretty sure that I've never seen it failing before.
As I mentioned in response to Jack's comment, that's now fixed.
The bug has been there for a long time, but since the test script uses
random matrices and doesn't set a seed, you could get lucky and have
eigenvalues with zero imaginary parts, in which case the script
wouldn't fail.
2. This gridplot example
https://sourceforge.net/p/gretl/workspace/ci/master/tree/tests/test_scrip...
also fails with the message:
<error>
*** error in function multiplot, line 32
> gnuplot --input="/home/artur/.gretl/outfile.3GVZ71"
--output="display"
called by function plot_season_plot
WARNING: Failed to execute file 'gridplot_seasonplot.inp'. Ignore.
</error>
That's inevitable I'm afraid. season_plot produces something that's
already a multiplot, and that cannot be handled by gpbuild (since
multiplots cannot be nested in gnuplot).
However... there's a fixable issue in the season_plot package which
would help. The package produces two plot components labeled
"obsmajor" and "obsminor" which can be plotted individually or
jointly. The joint plot is inherently a "multiplot" instance and so
can't work with gpbuild. The individual plots are NOT inherently
"multi" but the package writes them as if they were: the gnuplot
commands generated by season_plot look like this:
set multiplot layout 1,1 rowsfirst
# one individual plot
reset
unset multiplot
If the "set multiplot ", "reset" and "unset multiplot" lines
were
simply omitted, then a set of such plots could be handled by gpbuild.
Allin