Re: [Matlisp-devel] EIG problem

[Tunc S. <si...@ee...>]

FYI: after the recent (yesterdays) batch of fixes (thanks!) I did
a cvs update and now both Allegro and CMU on Solaris 
work fine w.r.t:

i) installation
ii) Michael's ZGEEV bug

Regards,
Tunc

Tunc Simsek wrote:
> 
> I just installed the latest cvs'ed matlisp on solaris.
> The install went find for CMUCL.  I saw earlier that
> you mentioned a package problem with SAVE-MATLISP.
> I did not get this; the saved image loads in a package
> called MATLISP-USER that is using the MATLISP package;
> I think that is ok.
> 
> The script below, however, does reproduce the error with CMUCL.
> 
> 
> I also installed using Allegro.  The install did not work
> because the f77-mangling.lisp and ffi-acl.lisp are not compatible.
> However, replacing all "_" with "" in f77-mangling is a sufficient
> hack to install using Allegro.  SAVE-MATLISP also works for Allegro.
> 
> The script below, again, does reproduce the error with Allegro.
> 
> MATLISP-USER(1): :ld test
> ; Loading /home/eclair3/varaiya/simsek/tmp/mk/test.lisp
>  ** On entry to ZGEEV  parameter number 10 had an illegal value
>  ** On entry to ZGEEV  parameter number 12 had an illegal value
> 
> Regards,
> Tunc
> 
> "Michael A. Koerber" wrote:
> >
> > The command (EIG A) seems to be okay, but not (EIG A T).  This
> > results in ZGEEV complaints as well as inconsistant error reports
> > for succesive calls.  A script demonstrating this follows...
> >
> > ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
> > ;; Demo a ZGEEV bug in MATLISP with
> > ;;
> > ;; CMU Common Lisp 18c, running on koerber.llan.ll.mit.edu
> > ;; Send questions and bug reports to your local CMU CL maintainer,
> > ;; or to cmu...@co.... and cmu...@co.... respectively.
> > ;; Loaded subsystems:
> > ;;     Python 1.0, target Intel x86
> > ;;     CLOS based on PCL version:  September 16 92 PCL (f)
> > ;;     MATLISP/Pre 2.0
> >
> > (use-package :m)
> >
> > ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
> > ;; Okay read in a test array from the data set (see bottom of file)
> > (with-open-file (in "tmp-ar.out" :direction :input)
> >                 (defparameter tst-array (read in)))
> >
> > ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
> > ;;;                         BEGIN THE DEMO
> > ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
> > ;; This first block runs through the computational process with no
> > ;; errors.  It is missing the MULTIPLE-VALUE-BIND blocks
> >
> > (setf tst-matrix (make-complex-matrix tst-array)
> >       tst-a (m* (ctranspose tst-matrix) tst-matrix))
> >
> > ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
> > ;; Finally note the error
> > (eig tst-a t)
> >
> > ;; ** On entry to ZGEEV  parameter number 10 had an illegal value
> > ;; ** On entry to ZGEEV  parameter number 12 had an illegal value
> >
> > ;;; Running this interactive will yeild different results...
> >
> > ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
> > ;;; THE DATA FILE...save this as "tmp-ar.out"
> > #2A((#C(16.005199412100566d0 47.07791625264504d0)
> >      #C(29.826457010016536d0 -35.54662313696197d0)

[snip]