swarm-support
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

About that "illegal instruction" error you see on SOME linux systems. (w


From: Paul E Johnson
Subject: About that "illegal instruction" error you see on SOME linux systems. (was Re: Swarm RH 8.0
Date: Fri, 29 Nov 2002 11:35:57 -0600
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020918

I've traced this problem back to some incredibly "fine grained" difference between CPUs and the compilation of the HDF5 library. I shipped a description to the hdf5 team, and they could not explain it either. They said hdf5's configure assumes you are on an i686 platform, but that could not explain the problem very well in my opinion, since most people who see this problem tell me they have Pentium III. It is easy to see that my precompiled hdf5 just won't work on a Pentium I or II machine, but PIII and IV should be OK. One might be concerned about some athlons. I'm beginning to suspect that this hdf5 problem is revealing some kind of "pentium bug". (Once I bought a Gateway pc and after a couple of years I learnt that MS Excel gave bad random numbers because of a flaw in the CPU that Intel acknowledged but did not fix).

Here is something you could try. Download the SRPM file for hdf5

http://lark.cc.ku.edu/~pauljohn/software/hdf5-1.4.4-1RH80.src.rpm

and see if you can rebuild it on your system. If you have the settings so that non-root people can build RPMS, then it should be as easy as "rpmbuild --rebuild blah.blah.src.rpm". It might be informative to install the SRPM, then try a build "rpmbuild -ba hdf5.spec" in the SPEC dir and, even if it fails, you will end up with the hdf5 build tree in your BUILD dir and you can do some experiments. If your system is like the other ones I've seen, the rebuild will fail and you will get some error messages we (me or you) could send to the hdf5 team.

If the build fails, you could look at the spec file in the SRPM and I think you could add a configure flag to force it to assume you are on i386.

I just learned of this problem after I built the hdf5 rpm, and I realize I'm lying when my rpm is packaged with the "i386" in the title when it ought to be i686.

If you want to create for me an account on your system and privately email me the name/password, I could dig around, but I'm almost certain (on the basis of experience with a few other users) that the problem will trace back to hdf5.

On the one other machine were I have been able to test this problem through, I was able to successfully build Swarm without the hdf5 support and that user was happy enough with that build.


Triola, Marc wrote:
Paul,
  I am a Medical Informatics fellow at New York University School of
Medicine.  One of my research projects involves simulating the
transmission of organisms in our Intensive Care Unit.  I have been using
the Easel programming language (www.cert.org/easel) with some success.
Because of the long time it takes to complete my sims, I am strongly
considering switching to swarm.  I am new to Linux and am running RH
8.0.  I downloaded your awesome rpm for 8.0 and installed it by double
clicking (seemed almost too easy).  When I try to make the heatbugs
files, I get the following output (below).  The subsequent executable
that is created just gives me an 'illegal instruction' error.  Any ideas
on what I've done wrong?
  Thanks,
      Marc

/bin/sh /usr/bin/libtool-swarm --mode link gcc -g -O2 -march=i386
-mcpu=i686 -L/usr/lib/swarm -rpath /usr/lib/swarm  -o heatbugs Heatbug.o
HeatSpace.o main.o HeatbugModelSwarm.o HeatbugObserverSwarm.o
HeatbugBatchSwarm.o  -lswarm -lobjc gcc -g -O2 -march=i386 -mcpu=i686 -o
heatbugs Heatbug.o HeatSpace.o main.o HeatbugModelSwarm.o
HeatbugObserverSwarm.o HeatbugBatchSwarm.o  -L/usr/lib/swarm
/usr/lib/swarm/libswarm.so
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/space
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/analysis
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/simtoolsgui
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/simtools
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/random
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/tkobjc
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/tclobjc
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/objectbase
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/activity
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/defobj
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/collections
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/src/misc
-L/home/pauljohn/LinuxDownloads/redhat/BUILD/swarm-2.1.141.20021019/=wit
h-hdf/libobjc -L/usr/lib -L/usr/X11R6/lib /usr/lib/swarm/libspace.so
/usr/lib/swarm/libanalysis.so /usr/lib/swarm/libsimtoolsgui.so
/usr/lib/swarm/libsimtools.so /usr/lib/swarm/librandom.so
/usr/lib/swarm/libtkobjc.so /usr/lib/swarm/libtclobjc.so
/usr/lib/swarm/libobjectbase.so /usr/lib/swarm/libactivity.so
/usr/lib/swarm/libdefobj.so /usr/lib/swarm/libcollections.so
/usr/lib/swarm/libmisc.so -lBLT24 -ltk8.3 -ltcl8.3 -lXpm -lpng
/usr/lib/libhdf5.so -lz -lX11 -lm -ldl /usr/lib/swarm/libobjc.so -lgcc
-Wl,--rpath -Wl,/usr/lib/swarm -Wl,--rpath -Wl,/usr/lib/swarm
-Wl,--rpath -Wl,/usr/lib -Wl,--rpath -Wl,/usr/X11R6/lib
_________________________________________ Marc M. Triola, M.D.
New York University School of Medicine
(212) 263-3978



--
Paul E. Johnson                       email: address@hidden
Dept. of Political Science            http://lark.cc.ku.edu/~pauljohn
1541 Lilac Lane, Rm 504
University of Kansas                  Office: (785) 864-9086
Lawrence, Kansas 66044-3177           FAX: (785) 864-5700


                 ==================================
  Swarm-Support is for discussion of the technical details of the day
  to day usage of Swarm.  For list administration needs (esp.
  [un]subscribing), please send a message to <address@hidden>
  with "help" in the body of the message.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]