🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: gnu.gcc.help
Article #50417

GCC on UNISYS 6075 (Sequent 750)

#50417
From: D. Doucet
Date: Tue, 28 Sep 1993 15:10
66 lines
3441 bytes

I have just spent a day building gcc-2.4.5 on a UNISYS 6075
(basically a sequent 750).  Seeing as there was no ready-made
config for ANY UNISYS box nor sequent-ptx, I had to go for the
generic i486-*-sysv32 configuration.  This compounded with the
fact that the site administrator wanted everything compiled in
the POSIX world lead to much dancing.

UNISYS' iAPX[34]86 native compiler is from mostly Sequent built
(the part that more-or-less works), however UNISYS has added
their own -j flag to signify that the POSIX world is to be used.
This didn't seem too bad at first --just give the -j to CFLAGS
when building stage1 and set the STANDARD_INCLUDE_DIR to /usr/include/posix
(where the POSIX headers are kept). Once stage1 is complete use
-Wl -j (or -Xlinker -j) to tell the native linker to use the
POSIX libraries and avoid all those 'bad magic numbers'.  

But, of course, that would be too easy.  It seems that many of
the POSIX header files just include the standard /usr/include
varity headers -- apparently UNISYS hack on the native cpp knows
enough to look in /usr/include for any nested includes from a
POSIX header as no relative paths are used: just <foo.h>.  
After succesfully throwing my newly built stage1/xgcc into a loop
and exhasting the virtual memory, I linked /usr/posix to
/usr/include/posix and set SYSTEM_INCLUDE_DIR to /usr/posix.
Now everthing went well, until I discovered that UNISYS failed to
provide anything but crt0.o with the POSIX libraries and the
whole process had to be aborted and (after explaining to the
Manager that it couldn't be done) we returned to the bcs world.

I removed all POSIX references and had to #define NO_SYS_SIGLIST
in the config file (I don't know why this wasn't a problem
before).  Now I have a problem with floating-point.  Mainly the
fact that enquire goes off to la-la-land when built with xgcc.
However, if I hardcode into the Makefile that the native compiler
be used to create enquire, it works and I get float.h (I still
haven't tested any floating point programmes with my new gcc, but 
I have this sinking feeling that it's not going to work).

Has anyone out there already made the port to either one of the
MP Unisys boxes, or to PTX, and can provide a set of
configuration files or hints?  Floating point is going to be
absolutely necessary.

Tommorrow, I have to build emacs-19.19, libg++-2.4 and glibc-1.06
on this box...so I am open to suggestions on these as well.

In (Pseudo)Reality,

Roy


Roy Nicholl

Atlantic Systems Group
Incutech Centre
B/S 69000
Fredericton, NB
Canada  E3B 6C2

phone	(506) 453-3505
fax	(506) 453-5004

email:	douc@jupiter.csd.unb.ca -or- 
	asg@jupiter.csd.unb.ca 
Technical Details
Message-ID: <1993Sep28.151024.27462@jupiter.sun.csd.unb.ca>
Path: arc1-ca.pugleaf.net!gmd.de!xlink.net!sol.ctr.columbia.edu!howland.reston.ans.net!usc!cs.utexas.edu!utnut!torn!csd.unb.ca!douc

Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 11318c,73e628f68be7be54,start
X-Google-Attributes: gid11318c,public
X-Google-Thread: 14ede5,73e628f68be7be54,start
X-Google-Attributes: gid14ede5,public
X-Google-ArrivalTime: 1993-09-28 08:30:27 PST
Newsgroups: gnu.gcc.help,gnu.gcc.bugs
Path: arc1-ca.pugleaf.net!gmd.de!xlink.net!sol.ctr.columbia.edu!howland.reston.ans.net!usc!cs.utexas.edu!utnut!torn!csd.unb.ca!douc
From: douc@jupiter.sun.csd.unb.ca (D. Doucet)
Subject: GCC on UNISYS 6075 (Sequent 750)
Message-ID: <1993Sep28.151024.27462@jupiter.sun.csd.unb.ca>
Organization: University of New Brunswick, Fredericton, NB, Canada
Distribution: gnu
Date: Tue, 28 Sep 1993 15:10:24 GMT
Lines: 66