Re: [Gcl-devel] Windows ANSI test - universe.lsp

[Camm Maguire]

Greetings!

Mike Thomas <address@hidden> writes:

> Hi all.
> 
> Dave and I have each had trouble in the ANSI test with universe.lsp on
> high optimisation.
> 
> By a process of elimination I found the site of the problem in the
> defparameter *characters* declaration at the error call which, when
> replaced with (print "error") (and supplemented with a print of the
> loop parameter) actually does print an error message rather than
> barfing:
> 
> ================================================
> (defun try-to-read-chars (&rest namelist)
>    (loop
>      for name in namelist append
>       (handler-case
>           (list (read-from-string
>                  (concatenate 'string "\#\\" name)))
>         (error () nil))))
> 
> (defparameter *characters*
>      (remove-duplicates
>       `(#\Newline
>         #\Space
>         ,@(try-to-read-chars "Rubout"
>                           "Page"
>                           "Tab"
>                           "Backspace"
>                           "Return"
>                           "Linefeed"
>                           "Null")
>         #\a #\A #\0 #\9 #\. #\( #\) #\[ #\]
>         )))
> ================================================
> 
> The error message:
> 
> ================================================
> Loading ansi-aux-macros.o
> start address -T 107f2680 Finished loading ansi-aux-macros.o
> Loading universe.lsp
> 
> "#\\Rubout"
> "#\\Page"
> "#\\Tab"
> "#\\Backspace"
> "#\\Return"
> "#\\Linefeed"
> "#\\Null"
> Error in READ-FROM-STRING [or a callee]: "NULL" is an illegal
> character name.
> ================================================
> 
> It's 9.30 at night and I've been going since 3.30 this morning (not
> all on programming of course) so I don't propose to chase this problem
> now or tomorrow, but if someone can take the baton it would be
> appreciated.
> 
> There are really two problems here:
> 
> 1. the Null character error message and

This should be there, but should be ignored by the handler case, and
probably is in fact but for the fact that it must go through FEerror
first where it finds the corruption.

Take care,


> 
> 2. the fact that error barfs in that location (it doesn't if just
> typed on the command line).
> 
> Cheers
> 
> Mike Thomas.
> 
> 
> 
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/gcl-devel
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah