testing Octave

[Toon Moene <toon at moene dot indiv dot nluug dot nl>]

>  There was some agreement previously that GNU Octave is
>  a good realistic test case.

Yep, this occurred to me as good testing case too (as would be  
Blitz++; unfortunately, they ask you to join their mailing list to  
obtain the sources, which seems a bit much for testing purposes  
only)

>  I've managed to test 2.0.13 on i586-pc-linux-gnu with
>  pre-egcs-1.1 using more aggressive Fortran options than
>  the default (-O2 -funroll-loops) and it passes (modulo
>  a known failure with mktime (?)).  This worked with the
>  latest dejagnu snapshot in infrastructure, whereas in
>  the past I had to use the released dejagnu for some
>  reason.

Well, I didn't even get this far.

>  In case anyone with plenty of cycles and bytes can to
>  try it on other systems, it turns out you need the patch
>  at
>  <URL:http://www.che.wisc.edu/octave/mailing-lists/bug-octave/1998/270>
>  for some C++-ism which egcs now rejects.

and not even that far:

gcc  -o info dir.o display.o doc.o echo_area.o filesys.o  
info-utils.o info.o infodoc.o infomap.o m-x.o nodes.o search.o  
session.o signals.o terminal.o tilde.o window.o indices.o xmalloc.o  
nodemenu.o footnotes.o dribble.o variables.o gc.o man.o clib.o  
getopt.o getopt1.o bzero.o  strdup.o -ltermcap
/bin/ld: Undefined symbols:
_tcgetattr
_tcsetattr
collect2: ld returned 1 exit status

BTW, this is for octave-2.0.9, the newest I could find on the gnu  
mirrors (but I used ftp.nluug.nl; perhaps they're left behind?).

Anyway, I am not going to pester the Octave author for what's  
probably an obscure autoconf misunderstanding of the support for  
termcap on NEXTSTEP3.2.

Alas,
Toon.