nsb@THUMPER.BELLCORE.COM (Nathaniel Borenstein) (02/23/90)
Well, I've finally gotten around to applying patch 3 to my version of Andrew, and it dies consistently in the build process on the Sun4. It is the usual kind of thing -- you type "make Install" in the atk/ness directory, and it comes back with the following: Checking Destination Directories.... touch install.time touch install.doc sh: 6141 Memory fault - core dumped *** Error code 139 make: Fatal error: Command failed for target `Install' All of the actual code seems to be compiling OK, and right now I've got the compilation moving along nicely by using "make -k Install". However, I hadn't seen this bug reported for patch 3 (though I had seen similar ones for earlier releases) and it never happened to me before, so I thought I should report it. -- Nathaniel
wjh+@ANDREW.CMU.EDU (Fred Hansen) (02/23/90)
Excerpts from mail: 23-Feb-90 Core dump in build Nathaniel Borenstein@thu (773+0) > Checking Destination Directories.... > touch install.time > touch install.doc > sh: 6141 Memory fault - core dumped > *** Error code 139 > make: Fatal error: Command failed for target `Install' Notice that the failure is in sh and not in ATK code. Jim Fulton and I spent two days during the integration process rearranging andrew.rls to get rid of these problems, but as the system changes, they reemerge. Generally the work-around is to simplify the instructions to sh found in andrew.rls. Since we haven't a SUN4 to play with here, maybe someone who has can post a fix. FredH PS. It would help to know exactly which directory the failure occurred in. Checking patches.003, however, I can find nothing that should have affected the sh commands generated in the Makefile. Sometimes the SUN4 sh failure is non-repeatable.
janssen@parc.xerox.com (Bill Janssen) (02/24/90)
Excerpts from andrew: 23-Feb-90 Re: Core dump in build Fred
Hansen@andrew.cmu.e (904+0)
> Sometimes the SUN4 sh failure is non-repeatable.
Yep, I compiled just fine with patch003 installed (Sun-4/260, SunOS
4.0.3, CDEBUGFLAGS = -g).
Bill
jw@SICS.SE (Johan Widen) (02/24/90)
>>>>> On Fri, 23 Feb 90 13:44:11 -0800 (PST), Bill Janssen <janssen@parc.xerox.com> said: > Sometimes the SUN4 sh failure is non-repeatable. janssen> Yep, I compiled just fine with patch003 installed (Sun-4/260, SunOS 4.0.3, janssen> CDEBUGFLAGS = -g). When I compiled andrew with cc -O on the Sun-4, I too got what at first seemed like a crash in sh. The problem turned out to lie in /bin/as rather than sh. Sun is apparently aware of this problem. I had no such troubles when I compiled andrew with gcc-1.37: gcc -O -traditional -fstrength-reduce -fwritable-strings (I used the gnu assembler gas-1.34). We have been running a gcc compiled version of andrew on our Sun-3's and SPARCstation's now for about a month without any apparent problems. Johan Widen SICS, PO Box 1263, S-164 28 KISTA, SWEDEN Internet: jw@sics.se Tel: +46 8 752 15 32 Ttx: 812 61 54 SICS S Fax: +46 8 751 72 30
nsb@THUMPER.BELLCORE.COM (Nathaniel Borenstein) (02/26/90)
Excerpts from info-andrew: 23-Feb-90 Re: Core dump in build Bill Janssen@parc.xerox. (240+0) > Excerpts from andrew: 23-Feb-90 Re: Core dump in build Fred > Hansen@andrew.cmu.e (904+0) >> Sometimes the SUN4 sh failure is non-repeatable. > Yep, I compiled just fine with patch003 installed (Sun-4/260, SunOS > 4.0.3, CDEBUGFLAGS = -g). > Bill Huh. Works for me now, too. How annoying. Sorry to bother people with it. -- NB