On 6/23/06, Ian Mclaren <mclareni@mail.cern.ch> wrote:
Dear Kevin, It is not clear from your convoluted log, but I suspect this is a 64-bit system and I have the same problem here, i.e. Comis does not work properly. Our 64-bit systems have i386 emulation and I have put statically linked ia32 versions of pawX11 and paw++ for them, which seem to work well.
No, this is a 32-bit system (and I've reproduced it on my own 32-bit system). I don't think this has to do with pointer size - rather, it seems to be an issue with the design of COMIS, that it first tries to use a PAW variable by trying to interpret it as the name of a function in any currently loaded libraries. The relevant code seems to be in the PAW source files pawfca.F, csaddr.F, and cs_hlshl.c (C function cs_get_func_()). I'm attaching a simpler test case that may better show what's going on. Untar the attached tarball and run "make" inside it. Then run paw from the same directory (so that it starts up by executing the enclosed pawlogon.kumac file). best regards, -- Kevin B. McCarty <kmccarty@princeton.edu> Physics Department WWW: http://www.princeton.edu/~kmccarty/ Princeton University GPG: public key ID 4F83C751 Princeton, NJ 08544
Attachment:
paw-test-case.tar.gz
Description: GNU Zip compressed data