[gnu.gcc] GNU CC on a stock ATT System V/386

cracraft@wheaties.ai.mit.edu (Stuart Cracraft) (04/10/89)

Hi,

This letter describes my experiences with GNU CC ("gcc")
on a stock ATT System V/386 machine (Toshiba 5100).

The system compiler is a pcc-derivative. Compilation
succeeded with the first Make (as described in INSTALL).

However, upon doing:

	make CC=stage1/gcc CFLAGS="-g -O -Bstage1/" c-parse.tab.c

... the make failed. The symptom of the failure is a lot
of disk accesses, maybe a lot of compiler temporary-file
work, and just an infinite suspension of belief.

This also happens when including "-msoft-float" in the
command line above. The same symtpoms occur (as described
in the previous paragraph). To the best of my knowledge, this
system does not have a floating point chip assist.

	Stuart

pcg@aber-cs.UUCP (Piercarlo Grandi) (04/15/89)

In article <1558@gluteus.ai.mit.edu> cracraft@wheaties.ai.mit.edu (Stuart Cracraft) writes:
    
    This letter describes my experiences with GNU CC ("gcc")
    on a stock ATT System V/386 machine (Toshiba 5100).

I am pleased. I have said that the T5100 and T5200 are the most comformtable
UNIX machines around, and apparently there are now some true believers. We
are now to the point that GCC is being installed on it.
    
    The system compiler is a pcc-derivative. Compilation succeeded with the
    first Make (as described in INSTALL).

    However, upon doing:
    
    	make CC=stage1/gcc CFLAGS="-g -O -Bstage1/" c-parse.tab.c
    
    ... the make failed. The symptom of the failure is a lot of disk
    accesses, maybe a lot of compiler temporary-file work, and just an
    infinite suspension of belief.

I have tried Enix 5.3.2, which is quite a stock AT&T 5.3.2, and it works
(except for the screen which is double spaced -- will have to set the video
modes to some funny number) on the T5100. SO I will tell you my hunch for
it. When memory is low, 5.3.2 simply locks up in the swapper, even if you
set the low water marks in stune that are supposed to stave deadlocks; it
often happens to me when I have several emacses. (I have only 2 megs).

The lot of disk accesses you hear is swapping... The problem does not occur
with cc because it is vastly smaller than gcc. Probably you are using a
stock T5100 with 2 megs. You can reclaim some space from the kernel,
otherwise, by deleting entries in sdevice (e.g., if you don't use
networking, streams modules like ptem, pts, timod, ....) and reducing the
size of many system tables in stune (you also need to modify mtune and
mdevice to change certain limits).

Details on application (I can send you my ?device and ?tune files).