Guetzlaff, Allan wrote:
> My attempts to build openssl-0.9.7l fail on ia64 hardware running HP-UX
> 11.23 (11i v2).
>
> Both the hpux-ia64-cc and hpux64-ia64-cc configurations fail when
> compiling bn_prime.c:
>
> cc -I.. -I../.. -I../../include -DOPENSSL_THREADS -D_REENTRANT
> -DDSO_DLFCN -DHAVE_DLFCN_H -DOPENSSL_NO_KRB5 -Ae +DD32 +O3 +Olit=all -z
> -DB_ENDIAN -DSHA1_ASM -c bn_prime.c
> Signal 11
> Signal 10
> cc: error 1922: "/opt/ansic/lbin/ctcom" terminated with signal 10
> *** Error exit code 1
>
> I am using HP's native compiler, but I was able to build openssl-0.9.7k
> without this error.
>
> Any suggestions on how to proceed with openssl-0.9.7l?


Looks like it got a signal 11 before it got a signal 10. IIRC that may mean it
ran out of memory (heap).

$ grep 11 /usr/include/sys/signal.h
* @(#)B.11.11_LR common/sys/signal.h $Revision: $
# define SIGSEGV 11 /* Segmentation violation */
# define SIGRTMAX ((int)___sysconf(10011)) /* _SC_SIGRTMAX */
# define SIG_HOLD ((void (*)(__harg)) (_NSIG+11))


The first bit of boiler-plate would be to make sure you are on the latest
patches for the compiler on the premis that the compiler got into a loop to
exhaust its memory. After that, see how close the compiler comes to hitting
maxdsiz (or the 64 bit version if file on /opt/ansic/lbin/ctcom shows it is
64-bit). Next, try dropping optimizations - first the +O3 I suspect, then
perhaps the +Olit.

Finally, generate a .i file - the compiler folks will want that to reproduce the
problem in-house.

rick jones
not a compiler guy

BTW, isn't 0.9.7 a bit old now? I thought that there were some perf
improvements for IA64 in later versions - 0.9.8 something or 0.9.9 - might be
worthwhile to try those...
__________________________________________________ ____________________
OpenSSL Project http://www.openssl.org
Development Mailing List openssl-dev@openssl.org
Automated List Manager majordomo@openssl.org