compiling bind 9.2.1 on HP-UX 11.0 problem

[Todd Bowden]

Im trying to compile bind 9.2.1 and I first compiled it under gcc 2.95. For the most part it was successful, except for the multiprocessor support. When I started up named I only saw one process and I think I should see 4 (being this a K460/4way), correct?

At this time I thought maybe it was GCC that didnt compile it correctly so Im trying to compile it with HP's ANSI C compiler but am getting the same error. I was wondering would anyone be able to help?

[/opt/bind-9.2.1] # make
making all in /opt/bind-9.2.1/make
making all in /opt/bind-9.2.1/lib
making all in /opt/bind-9.2.1/lib/isc
making all in /opt/bind-9.2.1/lib/isc/include
making all in /opt/bind-9.2.1/lib/isc/include/isc
making all in /opt/bind-9.2.1/lib/isc/unix
making all in /opt/bind-9.2.1/lib/isc/unix/include
making all in /opt/bind-9.2.1/lib/isc/unix/include/isc
cc -Ae -z -g -I/opt/bind-9.2.1 -I./include -I./../pthreads/include -I../include -I./../include -I./.. -D_REENTRANT +w1 +W 474,530 -c ipv6.c
cc -Ae -z -g -I/opt/bind-9.2.1 -I./include -I./../pthreads/include -I../include -I./../include -I./.. -D_REENTRANT +w1 +W 474,530 -c app.c
cc -Ae -z -g -I/opt/bind-9.2.1 -I./include -I./../pthreads/include -I../include -I./../include -I./.. -D_REENTRANT +w1 +W 474,530 -c dir.c
cc: "dir.c", line 167: warning 732: Different types treated as unsigned for >.
cc -Ae -z -g -I/opt/bind-9.2.1 -I./include -I./../pthreads/include -I../include -I./../include -I./.. -D_REENTRANT +w1 +W 474,530 -c entropy.c
cc: "../entropy.c", line 306: warning 732: Different types treated as unsigned for !=.
cc: "../entropy.c", line 746: error 1000: Unexpected symbol: "destroy_check".
cc: panic 2017: Cannot recover from earlier errors, terminating.
*** Error exit code 1

Stop.
*** Error exit code 1

Stop.
*** Error exit code 1

Stop.
*** Error exit code 1

Stop.

Why is it taking HP so long to come out with BIND 9.x on HP-UX 11.0 when they already have it for 11.i?

Thanks for any help in advance.

Todd Bowden

[Todd Bowden]

Just wondering if anyone had any ideas?

Thanks
Todd

[Andrew Cowan]

Hi Todd,

This is a definite programming problem. I saw a lot of these when I first started using the 64bit kernel arch.

As for multi-processor aware, my understanding is that the program becomes "thread aware" I.E. If one CPU does some work, then switches to another job, the next "least busy" CPU could pick-up the thread, and carry-on (seamlessly) processesing the requests. Thus the load is effectiveley balanced. If you have four slave processes (one per CPU) I would expect that you would soon get a situation where one had locked a file (or resource), then the others would have to wait until the lock was released, therefore performance would actually degrade.

[rick jones]

BIND 9's multiprocessing support is based on threading, not multiple discrete processes, so you would not see multiple processes.

I've never had that compile problem on 11. To compile on 11i requires the very latest release candidate of BIND9.

You might go looking into the .c file in question to see if there is something obvious there, or perhaps try a more recent download of the source bits.

[Todd Bowden]

To all,

I found I was lacking a couple patches for my ANSI/C compiler (Thank you HP). I then ran into a problem with the linker. That turned out to be a problem with the openssl version I had which was compiled by GCC and not the ANSI/C compiler. I got openssl 0.9.6g and compiled it with the ANSI/C compiler and did a remake on BIND and all is well.

Thanks to Mark Andrews, Internet Software Consortium, without his help I might still be lost today.