Conference smurf::buildhelp

2609.0. "tcr1 build problem still not resolved" by AOSG::FILTER (Automatic Posting Software - mail to flume::puck) Tue Oct 22 1996 12:27

Date Of Receipt: 	 7-OCT-1996 13:55:15.18
From: 	KAMLIA::useg_prc "USEG_PRC Review  07-Oct-1996 1352"
To: 	buildhelp@DEC:.zko.kamlia
CC: 	useg_reng@DEC:.zko.kamlia, overman@DEC:.zko.kamlia, rob@DEC:.zko.kamlia
Subj: 	tcr1 build problem still not resolved

Could someone from buildhelp please investigate the problem with the 
tcr1supportos build.

Reply to wolfson, cc to rob.

I'm attaching previous mail (sent to odehelp).

- Kathy



 =========================================================================

------- Forwarded Message

Return-Path: rob
Received: from flume.zk3.dec.com by kamlia.zk3.dec.com; 
(5.65v3.2/1.1.8.2/05Mar96-0145PM)
	id AA12559; Mon, 7 Oct 1996 11:39:12 -0400
Received: from puremagic.zk3.dec.com by flume.zk3.dec.com; 
(5.65v3.2/1.1.8.2/16Jan95-0946AM)
	id AA29806; Mon, 7 Oct 1996 11:39:10 -0400
Received: from localhost by puremagic.zk3.dec.com 
(5.65v3.2/1.1.10.5/09Aug96-0916AM)
	id AA09154; Mon, 7 Oct 1996 11:39:09 -0400
Message-Id: <[email protected]>
To: [email protected], duane
Cc: [email protected], useg_prc, [email protected]
Subject: Problems with tcr1supportos.nightly
Date: Mon, 07 Oct 96 11:39:09 -0400
From: rob
X-Mts: smtp

Hi,

Sorry, but I was out at a customer last week, and apparently noone has done
anything else with this...

I talked with Andy Duane who mentioned a problem with ld that can cause 
"hundreds" of "branch relocation out of range" errors if there is a multiply
defined, or undefined, symbol.  In looking at the logs for the failed build
on Sept 25, I noticed that there are two undefineds at the beginning of the
hard errors.  When I build GENERIC in my sandbox, I do not have this problem.
The only error message I get is in rm_spur where it does complain about the
branch relocation being out of range, but successfully builds a GENERIC kernel
with the trampoline code.

Here is the beginning of tcr1supportos.nightly/Logs/Error.summary:

kernel.bld.log:4569:Error: Undefined:
kernel.bld.log:4932:Error: Undefined:
kernel.bld.log:4449:cnx_hab_cfg.o: branch relocation out-of-range, bad object 
file produced, can't branch from 0xfffffc000064b8b4 to 0xfffffc000023f66c
kernel.bld.log:4572:mcs_asalloc.o: branch relocation out-of-range, bad object 
file produced, can't branch from 0xfffffc0000663184 to 0xfffffc00002450fc
.
.
.

I have no idea where the undefineds are coming from.  Could someone figure
out why we get them?  Again, I do NOT get these errors when I build the
kernel, so I suspect that a header file, or something, is being left off in
the build environment.

Also, tcr1supportos.nightly does not appear to have any cscope information.
Could someone please fix this?

Thank-you,

Rob Marshall
USEG

------- End of Forwarded Message
From:   ALPHA::"[email protected]" "26-Sep-1996 1551"
To:     wolfson@dec:.zko.smurf    (Kathy Wolfson, UNIX Sys. Release Eng.,DTN 381
-6108)
CC:     guru::odehelp, guru::rob, kamlia::useg_prc
Subj:   Re: TCR1supportos build errors
This isn't really an odehelp question since it has nothing to do with ODE. It is
a kernel developer question that the support engineer should work through his
organization or the kernel group.
In general this type of message occurs when the kernel grows too large to find
within the pre-defined size limit that existed at some point in the 3.2 world.
Since then Andrew Duane has introduced "trampoline code" that re does this
section and spits out a warning, but I believe that at this point in the life
span of the OSF kernel, the only thing you can do is make it small enough to
fit.
                                                              
        -Grant

Date Of Receipt: 	 7-OCT-1996 14:49:32.93
From: 	KAMLIA::shashi "Shashi Mangalat USG  07-Oct-1996 1447"
To: 	[email protected]
CC: 	[email protected], [email protected],
	[email protected], [email protected]
Subj: 	Re: tcr1 build problem still not resolved

I am confused!!

tcr1supportos.nightly was last built on Jun 22!  

	falpha [275] tail kernel.bld.log
	Relinking GENERIC with trampoline code
	chmod 755 a.out
	(null command)
	mv a.out vmunix.sys
	ln vmunix.sys vmunix; ln vmunix vmunix.GENERIC
	(null command)
	(null command)
	(null command)

	Elapsed build time(hr:min:sec): 3:2:41 
	falpha [276] ll kernel.bld.log
	-rw-r--r--   1 devbld   staff    1136932 Jun 22 08:23 kernel.bld.log

--shashi

Date Of Receipt: 	 7-OCT-1996 15:08:21.34
From: 	KAMLIA::rob "Rob Marshall USG  07-Oct-1996 1505"
To: 	[email protected]
CC: 	[email protected], [email protected],
	[email protected], [email protected]
Subj: 	Re: tcr1 build problem still not resolved

Hi,

In taking a brief look at this, and thanks to Grant's pointer, I have con-
firmed that the undefined symbols that seem to be causing this are in 
lockinfo.c.  Since noone has submitted any changes to tcr1supportos for
lockinfo.c, the question becomes: why is this now broken?  Have changes to
a pool to which tcr1supportos is backed caused this?  One of the symbols
is in a #if NFAA > 0, so it is possible that the value of NFAA has been
changed.  None-the-less this appears to be a build issue.

Thanks,

Rob Marshall
USEG