Re: Ebcdic rule

[Paul Eggert]

> From: Pfeiffer Daniel <address@hidden>
> Date: Tue, 2 Oct 2001 18:49:50 +0200
> 
> ever since everybody's shouting "We're open!", mainframes have learned a few
> Unix manners, which is a good thing.  Alas, while being quite close to
> Posix, they'll use Ebcdic rather than Ascii, which causes no end of
> headaches.

Wow.  Are you actually using such mainframes to run GNU applications?
I thought that, these days, most of those people were running in ASCII
(or Latin-1 or UTF-8) mode, even on mainframes.


> [ "`echo A`" = "`echo '\0301'`" ]

Why do you need an Autoconf test for this?  Can't you put
something like the following into your C code?

#if 'A' == '\301'
 printf ("Looks like we're using EBCDIC.");
#endif

Also, some 'echo' implementations don't treat backslash specially,
so your shell script looks fishy.


In my experience, the main problem with porting to EBCDIC is not
detecting whether the host uses EBCDIC; it's all the existing code
that assumes ASCII without thinking.  E.g., 'a' <=x && x <= 'z'
succeeds for some bytes that are not lower-case letters in EBCDIC.

(And don't get me started about east-Asian EBCDIC extensions like
EBCDIK.  :-)