[Simh] [simhv37-3] gcc-2.95.3 and OpenBSD 3.x

Nelson H. F. Beebe beebe at math.utah.edu
Fri Mar 14 19:38:28 EDT 2008


I've been porting a large mathematical library that I'm developing to
the VAX platform provided by simhv37-3 with OpenBSD versions 3.5 and
3.8, and I've run into a serious gcc problem that can be exhibited by
a single-line test file:

    % gcc --version
    2.95.3

    % cat bug3.c
    static double x = 366503875925. / ( 1099511627776.0 * 1099511627776.0 * 1099511627776.0 * 1099511627776.0 );

    % gcc -c bug3.c
    gcc: Internal compiler error: program cc1 got fatal signal 11

The internal error happens under ALL optimization levels and on both
versions of OpenBSD.  If the number of constants in the denominator is
reduced from four to three, the code compiles.  It also compiles if
the constants are reduced in size (e.g., 1.0 / (2.0 * 3.0 * 4.0 *
5.0)).

Curiously, the same compiler version on my NetBSD 3.0 SIMH VAX system
compiles this code without problems.

This is a gcc and/or OpenBSD problem, not a simh one.  Nevertheless,
has anyone on this list successfully built a later version of gcc for
OpenBSD VAX than gcc-2.95.3, and if so, for what version?  Is there a
binary distribution of that compiler build available?

I've done more than 1100 gcc builds on more than 20 flavors of Unix,
and based on that experience, I have found that most gcc-3.x and
gcc-4.x builds fail almost everywhere.  On the SIMH VAX, a single
build of those versions of gcc could take several days, so it is not
easy to make trial builds.

-------------------------------------------------------------------------------
- Nelson H. F. Beebe                    Tel: +1 801 581 5254                  -
- University of Utah                    FAX: +1 801 581 4148                  -
- Department of Mathematics, 110 LCB    Internet e-mail: beebe at math.utah.edu  -
- 155 S 1400 E RM 233                       beebe at acm.org  beebe at computer.org -
- Salt Lake City, UT 84112-0090, USA    URL: http://www.math.utah.edu/~beebe/ -
-------------------------------------------------------------------------------



More information about the Simh mailing list