fmbutt@mrbt.sw.stratus.com (Farooq Butt) (03/07/91)
Hi there net.c++.brains: I need your help in figuring out some weird stuff I am seeing when trying to run programs linked to NIH 3.0. I am running on a big-endian machine, BTW and have ported AT&T cfront 2.1 OK. The problems arise in the initialization sections of NIH (i.e. nothing works). The entire NIH lib compiles beautifully into a large libnihcl.a archive and I get no link errors when linking programs with it. I have compiled NIH with -DDEBUG defined so it emits diagnostics while running. Here's a trace of my attempt to run "array.c" in the "test" directory: Class[Collection ident: Foo Port II version: 0 Memory fault(coredump) # rm core # sdb ./array no process *t no process *r ./array Class[Collection ident: Foo Port II version: 0 SIGNALED 11 process 20141 function computeSignature__5ClassFv() in Class..c 252: <no source text available> *t computeSignature__5ClassFv(__0this=0x35bd0) [Class..c:252] computeSignature__5ClassFv(__0this=0x30174) [Class..c:252] dumpOn__5ClassCFR7ostream(__0this=0x30174,__0strm=0x2d440) [Class..c:220] initialize__5ClassSFv() [Class..c:343] initialize__5NIHCLSFv() [Nil.c:175] _main() [_main.c:58] main() [array.c:38] _start() # Have any of you out there seen any problems like this ? Is NIH horribly endian-sensitive (my guess) ? Is this a bug ? I would assume that *somebody* out there in net.land has run into this little gem by now... Any scraps you throw my way will be hungrily consumed :-) Thanks fmb -- Standard High-Tech Disclaimer: NOTHING in the above article has the slightest relationship to reality. If any reality correspondences are found, please notify me IMMEDIATELY. Any threats, abuse or stupidity of any kind is purely UNintentional. These are MY Opinions NOT my employer's.