[Groff] Re: groff ported to OS/390

[Werner LEMBERG]

> I want to let you know I have ported groff in the last days to
> OS/390.

This is good news!

> The main problem in porting is that the IBM Mainframe machine uses
> EBCDIC internally, which is a real 8bit character set.
>
> [...]

Sándor,


your mail arrives exactly at the right time.  I'm going to abandon the
input encoding restrictions completely.

Please subscribe to address@hidden and download the current
development snapshot

  ftp://ftp.ffii.org/pub/groff/devel/groff-current.tar.gz

Some comments to your patches -- which I won't apply immediately since
I believe that we need more discussion how to solve the whole issue
gracefully (not to mention support for CJK encodings and Unicode which
is a different pair of shoes, so to say).

> ! DEFS=`sed -f conftest.defs confdefs.h | tr '\n' ' '`

Do all UNIX versions of `tr' support the `\n' notation?

>   char illegal_char_table[256]= {

My idea is to activate illegal characters only incompatibility mode --
BTW, does a UNIX troff exist on the OS-390?

> ! const int ESCAPE_CIRCUMFLEX = 010;
> ! const int ESCAPE_LEFT_BRACE = 011;

Again, I think this should be moved to values larger than 0xFF...

> !     fputc(had_prompt ? ':' : '\057', stderr);

For such situations I would like to have a config file, say,
groff/arch/i390/config.h -- I'll check how other packages manage this.

> I deleted the changes to the font description files - they must be
> completely replaced due to EBCDIC coding.  It would be a nice
> excercise for the autoconf step (if host=i390-ibm-mvs), include an
> other set of PROTO files and define the IS_EBCDIC_HOST macro.

If you provide PROTO files this should not be too difficult.


    Werner