help with compiling and running the X11R4 VGA server, Take 2

[Pim Zandbergen]

Did anyone succeed in compiling the X11R4 VGA Take 2 server 
for ISC yet?

I have the binary distribution up and running like a charm.
It is *much* faster than the previous server, even though that
was a 16 color server.  Thanks a bunch to Thomas Roell!

The only problem I had was that I had to recompile the 75dpi
and the 100dpi fonts. You can't use the ones from the old server.


For those who are interested, I use a Taxan 585 VGA card,
which is an OEM version of the STB VGA EM-16.

It has 512 kB RAM, the original STB ROM, a Tseng ET3000AX chip
and oscillators (according to what is printed on the parts) of
28.322, 37.25, 25.175, 44.9 and 65.0 MHz.

My Monitor is a standard, non-multisync VGA monitor.

This is what I specified in Xconfig:

# chip   ram   virtual   clocks                     default-mode  flags
ET3000   512   864 606   25 28  0 45  0  0 39       "640x480"   # STB VGA EM 16

# name       clock  horzontal timing      vertical timing      flags
"640x480"   25     640  672  768  800    480  490  492  525

I'd like to find out more of the possibilities. Unfortunately,
I don't have a Turbo Pascal compiler to compile the test program.
Could anyone post the compiled Pascal test program ?



The binary distribution contains a server that does not support
TCP/IP connections. According to the docs, the sources do.

I would like to have TCP/IP support for connecting to
binary-only clients that were written for ISC's X11 (will this work ?),
and eventually for X-terminals.

The sources however, give a lot more problems than I had with
the previous server. Some of them may be related to the fact
that I am still running ISC 2.0.2 with TCP/IP 1.1.1.
I do have the upgrades, but have not found time to upgrade yet.

>From memory, these are some of the problems I encountered:

1. Multiple inclusion of <sys/fcntl.h> in lots of sources.

   My solution: fix the header file to protect it against
   multiple inclusion.


2. mit/server/ddx/at386/vga/et[34]000/driver.c would not compile.
   I use gcc/gas with Michael Bloom's COFF patches. gas will not
   assemble gcc's output and nor will as.

   My solution: use cc instead


3. gcc barfs on the assembler statements in <netinet/in.h>

   My solution: add -DNO_ASM to the CFLAGS


4. mit/server/os/bsdemul/Berklib.c would not compile

   My solution: undefine insque() and remque()
   as macro's before declaring them as functions.


5. linker problems with shared libraries.
   This is the first of many similar messages
   from the output of 'make World' (long lines folded):

	making all in ./extensions/test...

	[ ... ]

	gcc -DNOSTDHDRS -DNO_ASM -fstrength-reduce -fpcc-struct-return \
	-fwritable-strings -o shapetest shapetest.o -O \
	../.././extensions/lib/libXext_s.a  ../.././lib/X/libX11_s.a -lpt -lc_s

	ld fatal: cannot combine NOLOAD input sections with other input sections;
	possibly caused by shared library address space overlap
	*** Error code 1

   My solution: beats me!

Thanks in advance for all your help and suggestions.
-- 
Pim Zandbergen                          domain : pim at cti-software.nl
CTI Software BV                         uucp   : uunet!mcsun!hp4nl!ctisbv!pim
Laan Copes van Cattenburch 70           phone  : +31 70 3542302
2585 GD The Hague, The Netherlands      fax    : +31 70 3512837