From: Michael Castle <mcastle AT umr DOT edu>
Subject: Re: PBM stuff?
To: dridge AT MIT DOT EDU (Matthew Eldridge)
Date: Mon, 9 May 1994 13:36:25 -0600 (CDT)
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu

Amazingly enough Matthew Eldridge said:
> alright, I give up.  What is the "NetPBM" distribution?  I've heard
> it tossed around before.  Is it a rework of Jef's original PBM suite?

Kinda.  Basically it's a group of people headed by (I *think*)
Bill Davidsen (he used to moderate comp.binaries.ibm.pc and a 386
mail based discussion list).  They basically tried to gather up
all the bug fixes and new p?m programs that others have written,
and packaged them all together.  

Jef's not part of it, but he is aware of it.  And true to the
nature of those involved, tempers often flare about how things
should be done.  Quite amusing really.  :->

Source can be picked up at ftp.wustl.edu in
graphics/graphics/packages/NetPBM
or something like that.... the first 3 subdirs are right
anyway...


If you apply the patches that I included with my PBMPlus port,
and do any other fixups (namely editing the other program names
in the p?mmerge.c files), it should compile fairly easily (I
would recommend using a make that swaps itself out of memory as
it does recursive makes, and you'll quickly run out of go32 room
with the current djgpp based make).  

When I did the port, I cheated most heinously.  I used a modified
crt0.o startup file.  *hangs head in shame*  Problem is, of
course, that stdin and stdout need to be in binary mode.  So, I
used a modified crt0.s that merely jumped to a void f(void)
function I had added to libc.a that did setmode()s as appropriate
(calling this function just before it called main()).  It seemed
to work quite well, but it is really a gross hack.  Still...
better than modifiying 50+ *.c files, specially when I didn't
know sed/awk/perl/etc at the time and had to do it by hand (not
that i really know them know ;-).

Next time I do it though, I will probably follow the scheme that
was used in the os/2 port. 

The os/2 port used ibm's c-set/2 compiler, and so didn't have
access to nor the ability to modify the crt0.obj, so he had to
come up with another method.

Basically, he put a small main() at the end of the on *.h file
included in every *.c file.  This main did the setmode()s, then
called next_main(argc,argv).  Then after that, he had the
following:

#define main next_main

So that the real main in the *.c file got renamed.  This will
require a bit more work to work with merged files (mulitple
definitions of next_main), but I think I can work around that.

Just some objservations on porting things using djgpp.

cheers
mrc

-- 
Mike Castle .-=NEXUS=-.  Life is like a clock:  You can work constantly
  mcastle AT cs DOT umr DOT edu     and be right all the time, or not work at all
S087891 AT UMRVMA DOT UMR DOT EDU   and be right at least twice a day.  -- mrc
    We are all of us living in the shadow of Manhattan.  -- Watchmen