Well, GCL 2.6.7 seems to be the current production version, so I think we
should certainly not have test suite failures under it. As to the rest,
CLISP 2.41 is not a particularly old version (e.g., it is the version
included in current Slackware distributions) so if it is not a big effort,
we should be compatible with it, too. In any case, these test suite failures
may be just symptoms, and the underlying actual problems may not be specific
to the lisp version.
I am worried about the expected errors in the test suite, because I fear
that they may undermine trust in the correctness of results produced by
Maxima. I think it is important to take these test suite failures seriously.
Viktor
-----Original Message-----
From: Raymond Toy [mailto:raymond.toy at ericsson.com]
Sent: Wednesday, November 19, 2008 7:45 AM
To: Viktor T. Toth
Cc: maxima at math.utexas.edu
Subject: Re: [Maxima] Test suite failures
Viktor T. Toth wrote:
> I am getting some test suite failures after a full recompile.
>
>
> 1. In the main test suite, I get
>
> a. tests/rtest14.mac, problems: (231 232 242) under CLISP 2.41
These all work for me with clisp 2.47 (and 2.44.1 previously, I think.)
>
> b. tests/rtest16.mac, problems: (110 118 119) under GCL 2.6.7
>
These all work with gcl 2.6.8pre, whatever that is.
> c. tests/rtest4.mac, problem: (86) under CMUCL 19c
> (This may be a known problem due to the old version of CMUCL I am using.)
These all work for me with CMUCL 2008-11.
>
>
> 2. The share testsuite also fails:
>
> a. share/contrib/stringproc/rtestprintf.mac, problems: (7 27 34 35) under
> CLISP 2.41
>
> b. rtest_pdiff, problem: (error break), and
> $RTESTPRINTF, problem: (error break) under GCL 2.6.7
> (Scrolling back, problem 82 in rtest_fourier_elim was correct despite an
> expected failure, while problems 7, 29, and 33 failed in rtestprintf.mac.
> However, rtestprintf caused an error break: $RTESTPRINTF, which is
probably
> what interfered with the final statistics.)
>
> c. $RTEST_ROMBERG, problem: (error break), and
> share/contrib/stringproc/rtestprintf.mac, problems: (7 34 35) under
CMUCL
> 19c
I get many more failures with CMUCL. But I think most of them are due
to different expectations. For example, some tests expect results like
1.2E+3, but CMUCL uses a lower case e. Some are questionable because of
the number of digits printed.
>
> I also noticed what appears to be an increase in expected errors. Are
these
> really harmless or are real problems being swept under the rug?
I think they are real problems. I think some have been there since the
beginning and have never been fixed. Some are newer, representing real
bugs.
Ray