data misaligment

lukef

unread,

Mar 22, 2012, 1:30:23 PM3/22/12

to

hi, having some issues locating the offending data item, code
location:

12-03-12 09:48:50.090 \HAL.2,760 TANDEM.GUARDLIB.D40 108
Misalignment Record Collected for Processor 2, PIN
701
Process Name: \HAL.$A2:193982501
Program Name: \HAL.$OBJ.OBJECTS.AUTH
Exception Count: 979220
Reference Address: 0x00005D73
Emulation Mode: Accelerated
Priv Mode: Non-Priv
PC: 0x704828D0
Code Space: (acc UC)
Object Name: (See Program Name)

what's the easiest way to locate the variable associated with the ref
address and how do i identify the line of code being executed? i
don't really get the PC value...

thanks & best regards, luke.

dimandja

unread,

Mar 22, 2012, 2:05:07 PM3/22/12

to

Compile your code with INNERLIST. Generated addresses should point
you to the line of code..

lukef

unread,

Mar 23, 2012, 5:56:03 AM3/23/12

to

On Mar 22, 6:05 pm, dimandja <demou...@yahoo.com> wrote:
> Compile your code with INNERLIST. Generated addresses should point
> you to the line of code..

hi thanks for the reply and i will give it a go. the problem i am
having is the code address (PC) in the ems event is so large
0x704828D0.
i think this is a millicode address ?
i can't use the low level inspect command (=) i would usually use to
identify the offending line of code.

dimandja

unread,

Mar 23, 2012, 11:32:02 AM3/23/12

to

I think you're right.

Try recompiling without acceleration and see if you run into the same
problem.

dimandja

unread,

Mar 23, 2012, 8:44:10 PM3/23/12

to

Luke,

I have some more time to think about this. Although I'm not familiar
with the environment you are dealing with, I'll try to give you some
pointers, until of course someone who understands your problem comes
along.

Lets start with some basic stuff.

It looks like a called routine is reporting this error. If that's the
case, it should be simple enough to breakpoint before (or after) the
call. Examine the buffer for misaligned data.

If your are sending data to other programs, you may try to use the
'align' command when you compile.

See if you can compile everything with symbols (and Innerlist) and
without acceleration, before debugging.

Let us know what your environment is (OS, language, modules called,
libraries used).

MicroTech

unread,

Mar 26, 2012, 12:41:16 AM3/26/12

to

On Friday, March 23, 2012 1:30:23 AM UTC+8, lukef wrote:

>> i don't really get the PC value...

Worry not, Luke: region 70 is UC (User Code), the millicode region starts higher up, at address 7E000000.

For a complete map of the fixed region addresses (code and data), please see GAME (www.tinyurl.com/tandem-game), Session 3, Slide 4.

Cheers,

Henry Norman
MicroTech Consulting
sites.google.com/site/microtechnonstop

lukef

unread,

Mar 26, 2012, 5:42:06 AM3/26/12

to

@dimandja The program is a BASE24 classic ATM AUTH process - a TAL
program that uses the XPNET SKEL library. O/S is G6.32.
unfortunately i cannot recreate the problem in the test environment so
can't debug to find the problem, i only have the EMS event to go on.

@henry, thanks for that but unfortunately for some reason your url is
blocked here. is there an easy way if identifying the code called?

Tone

unread,

Mar 26, 2012, 7:18:33 AM3/26/12

to

Try using APE (Accelerated Program Examiner). For example :

APE
>file AUTH
>procs tns/r - will show entry addresses
>icode at <proc> tns/r - will display instructions

Further info can be found in $SYSTEM.SYSTEM.APEDOC

dimandja

unread,

Mar 26, 2012, 9:25:01 AM3/26/12

to

> @henry, thanks for that but unfortunately for some reason your url is
> blocked here.

That URL will not work on groups.google.com. You must append "http://"
like this:
HTTP://www.tinyurl.com/tandem-game

The actual URL Henry is pointing you to resolves to:
http://www.box.com/shared/bzz23tb2gu

dimandja

unread,

Mar 26, 2012, 9:40:29 AM3/26/12

to

> @dimandja The program is a BASE24 classic ATM AUTH process - a TAL
> program that uses the XPNET SKEL library.

I am no longer familiar with the sequence of calls between your
processes. I'll rely on your input while I try to be helpful.

1. How often does this problem occur?
2. Did you recently install a new module?
3. Did you recently make changes to some module?

I think the routine that reports the error is really not the one that
created the problem. You need to look upstream. You may have a user
written program (or procedure) that is called only on certain
occasions that coincide with the frequency of your error report.

If you have a mixed-language environment, you may want to use field
alignment commands such as the ones in DDL.

lukef

unread,

Mar 26, 2012, 12:40:31 PM3/26/12

to

thanks all, think it's sorted using APE. as i understand it this
allowed me to map the PC value from the EMS message (the accelerated
millicode address?) to the TNS address that INSPECT understood:

LOOKING FOR:
PC: 0x704828D0

> APE
APE (Accelerated Program Examiner) - T9292D30 - 13JUN05 (May 14 2005)
Copyright Tandem Computers Incorporated 1991-1994
APE> file auth4q
opened.
APE> procs tns/r
Base Length Entry Name
…
…
0x704825BC 0x00000688 0x70482B60 CSM074^CHECK^PBF^LIMITS
…
…
APE> icode at CSM074^CHECK^PBF^LIMITS tns/r
Initializing tables for code segment 0
EPIL table loaded
EPFL table loaded
All tables loaded

0x704825BC -> 0x70482C40 CSM074^CHECK^PBF^LIMITS
…
…
0x704828B0: ADD $0,sp,at LHU s0,-16(fp) LH
s4,18(fp)
0x704828BC: SLL t0,s0,1 LWL t5,0(t0) LWR
t5,3(t0)
0x704828C8: LHU s5,20(fp) ADDI t0,t5,262 LH
s0,0(t0)
0x704828D4: LHU s1,2(t0) LHU s2,4(t0) LHU
s3,6(t0)
0x704828E0: LHU s6,22(fp) LHU s7,24(fp) SUB
t7,s0,s4
0x704828EC: BNE t7,$0,0x7048291 NOP SUB
t7,s1,s5
0x704828F8: BNE t7,$0,0x7048291 NOP SUB
t7,s2,s6
…
…
APE> icode at CSM074^CHECK^PBF^LIMITS BOTH
Sp Pep Base Limit Entry Attrib Name
00 076 100000 100344 100265 CSM074^CHECK^PBF^LIMITS
…
…
100135:000135: 1 DADD 0x704828C8: LHU s5,20(fp)
100136:000136: 3 LQX 0x704828CC: ADDI t0,t5,262
100137:000137: 4 LADR L+011 0x704828D0: LH s0,0(t0)
100140:000140: 7 QLD 0x704828D4: LHU s1,2(t0)
100141:000141: 7 QCMP 0x704828D8: LHU s2,4(t0)
100142:000142: 7 BNEQ +010 0x704828DC: LHU s3,6(t0)
0x704828E0: LHU s6,22(fp)
0x704828E4: LHU s7,24(fp)
0x704828E8: SUB t7,s0,s4
0x704828EC: BNE t7,$0,0x7048291
0x704828F0: NOP
0x704828F4: SUB t7,s1,s5
0x704828F8: BNE t7,$0,0x7048291
0x704828FC: NOP
0x70482900: SUB t7,s2,s6
0x70482904: BNE t7,$0,0x7048291
0x70482908: NOP
0x7048290C: SUB t7,s3,s7
0x70482910: BNE t7,$0,0x7048292
0x70482914: NOP
…
…
> RUND auth4q
INSPECT - Symbolic Debugger - T9673G10 - (21JUL2008) System \HAL
Copyright Hewlett Packard Company 1983,1985-1999,2000-2005
INSPECT
003,00,00612 $Y668 #AUTHS^LN^04.#3025(AUTH4QS) + %101I
Executing command file INSPCSTM
SOURCE mode has been enabled
-$Y668-LOW
_$Y668_=100137:UC.00
= #CSM074^CHECK^PBF^LIMITS.16 + %7I,
#CSM074^CHECK^PBF^LIMITS.#29625(AUTH4QS) + %7I,
#CSM074^CHECK^PBF^LIMITS + %137I
_$Y668_HIGH
-$Y668-SO AT #CSM074^CHECK^PBF^LIMITS.#29625
#29621 !
#29622
#29623 ! Allow for possibility that PBF server has updated
the recor
#29624 ! If max bal is ascii spaces, set it to zero
@#29625 IF PBF1.BASE.USER^FLD^ATMOS.MAX^BALANCE =
ASCII^SPACES^F THEN
#29626 PBF1.BASE.USER^FLD^ATMOS.MAX^BALANCE := 0F;
#29627
#29628 IF PBF2.BASE.USER^FLD^ATMOS.MAX^BALANCE > 0F THEN
#29629 MAX^BAL := PBF1.BASE.USER^FLD^ATMOS.MAX^BALANCE
#29630 ELSE
-$Y668-

wbreidbach

unread,

Mar 28, 2012, 9:23:17 AM3/28/12

to

There could be a simple reason for that. The fieldalignment might be different using the accelerator. I had some interesting concerning that on Integrity.
Just define the structures:
struct .mystruct FIELDALIGN (SHARED2);
Maybe that will do the trick.

MicroTech

unread,

Mar 30, 2012, 10:13:56 PM3/30/12

to

On Monday, March 26, 2012 9:25:01 PM UTC+8, dimandja wrote:

> That URL will not work on groups.google.com. You must append "http://"

Many thanks, Dimandja! Strange, I use Google groups, works fine here... Could it be browser dependent? I'm a Chromer...

dimandja

unread,

Mar 31, 2012, 9:09:26 AM3/31/12

to

I use Chrome or Safari. Right now I'm using Chrome 18.0.1025.142 m.
Let's see how the URLs below render:

hp.com
www.hp.com
http://hp.com

The above all resolve to www.hp.com on Google Chrome.

dimandja

unread,

Mar 31, 2012, 9:11:33 AM3/31/12

to

On Mar 31, 9:09 am, dimandja <demou...@yahoo.com> wrote:
> On Mar 30, 10:13 pm, MicroTech <Henry.KO.Nor...@gmail.com> wrote:
>
> > On Monday, March 26, 2012 9:25:01 PM UTC+8, dimandja wrote:
> > > That URL will not work on groups.google.com. You must append "http://"
>
> > Many thanks, Dimandja! Strange, I use Google groups, works fine here... Could it be browser dependent? I'm a Chromer...
>
> I use Chrome or Safari. Right now I'm using Chrome 18.0.1025.142 m.
> Let's see how the URLs below render:
>

> hp.comwww.hp.comhttp://hp.com

>
> The above all resolve towww.hp.comon Google Chrome.

From groups.google.com, only htpp://www.hp.com works. Strange.

dimandja

unread,

Mar 31, 2012, 6:05:17 PM3/31/12

to

On Mar 31, 9:11 am, dimandja <demou...@yahoo.com> wrote:
> On Mar 31, 9:09 am, dimandja <demou...@yahoo.com> wrote:
>
> > On Mar 30, 10:13 pm, MicroTech <Henry.KO.Nor...@gmail.com> wrote:
>
> > > On Monday, March 26, 2012 9:25:01 PM UTC+8, dimandja wrote:
> > > > That URL will not work on groups.google.com. You must append "http://"
>
> > > Many thanks, Dimandja! Strange, I use Google groups, works fine here... Could it be browser dependent? I'm a Chromer...
>
> > I use Chrome or Safari. Right now I'm using Chrome 18.0.1025.142 m.
> > Let's see how the URLs below render:
>
> > hp.comwww.hp.comhttp://hp.com
>

> > The above all resolve towww.hp.comonGoogle Chrome.

>
> From groups.google.com, only htpp://www.hp.comworks. Strange.

Google Group URL resolution problem solved. By Google.

You must use the new groups.google.com though. When logging into (the
old) groups.google.com/group/comp.sys.tandem, you'll see a message at
the top of the page that says, "The old Google Groups will be going
away soon. Switch to the new Google Groups". Clicking on that link
will take you to the new stuff that does recognize all forms of URLs.

However, I must say, may brief test results of the new Google Groups
was a bit hairy and too sharp on the edges for my taste... I went
straight back to the old version...