Compiler bug in egcs-1.1

David Ronis ronis@onsager.chem.mcgill.ca
Wed Nov 11 21:40:00 GMT 1998


Here is a full (I hope) bug-report of what I posted to the egcs-bugs list.

I've been trying to compile dosemu-0.99.3.  I compile on an i586
running linux-2.0.35 and on an i486 running the same OS.  Both
machines run

	 gcc version egcs-2.91.57 19980901 (egcs-1.1 release)

The build was configured with the default options on both machines.
The compilation fails when trying to build int.c (appended below).  The
command (with -save-temps added by me) used is:

gcc -save-temps -c -O2 -fomit-frame-pointer -m486 -malign-loops=2 
-malign-jumps=2 -malign-functions=2 -MD -pipe -DUSING_NET -Wall 
-I/home/ronis/dosemu-0.99.3/src/include 
-I/home/ronis/dosemu-0.99.3/src/dosext/dpmi 
-I/home/ronis/dosemu-0.99.3/src/include/slang -I/usr/X11R6/include -o int.o 
int.c

gcc: Warning: -pipe ignored since -save-temps specified
gcc: Internal compiler error: program cc1 got fatal signal 11

The problem is reproducible on both machines.  In addition, if I reduce the 
optimization to -O1 the problem disappears (again onn both machines).

int.i is about 145K uncompressed and still too large for the list even when 
compressed.  I've put it on my web server at:

http://ronispc.chem.mcgill.ca/ronis/int.i

I built an unmodified egcs, and although I no longer have the source files and
cannot tell you what configure options were used,  I did a full bootstrap 
install, including the compareison of stage2 and stage3 compilers.  Most likely, 
I upped the optimization options slightly, but nothing else.

David Ronis





More information about the Gcc-bugs mailing list