Delivery-Date: Sat, 19 Mar 94 18:38:34 -0800
Message-Id: <199403200238.SAA19562@nobozo.CS.Berkeley.EDU>
To: aoki@postgres.Berkeley.EDU (Paul M. Aoki)
Subject: Re: more regression test info re:hpux port 
In-Reply-To: Your message of "Sat, 19 Mar 1994 18:28:46 MST."
             <199403200228.SAA27769@faerie.CS.Berkeley.EDU> 
Date: Sat, 19 Mar 1994 19:38:27 -0700
From: Daniel Simmons <simmdan@kenya.isu.edu>


In message <199403200228.SAA27769@faerie.CS.Berkeley.EDU> you write:
>> Well, I decided that it was pretty stupid of me to not just change
>> those permissions and try it.  So, I did and the regression tests
>> went to completion.
>
>:-)  thanks.  whew.  you just lowered my stress level by about
>an order of magnitude.

Good.  I guess I would mention that some thought ought to be given to
why the permissions were wrong in the first place, though.  Maybe it
is all just the fault of bsdinst, but I don't know.

>everything looks ok (unimportant diffs) except:
>
>< |             | -34.84      | -3.26607421344208|
>> |             | -34.84      | 0           |
>
>not good.  probably some random difference between the hpux 9.00 math 
>library (for 800s) and the 9.01 math library (for 700s).  i'll have to 
>check it out a 700 to make sure.  (i generally test on a 9000/877 
>running 9.00.)

Well, in the default configurations the 700s have math coprocessors
while the 800s do not...  One thing you have going for you, though, is
that the 877 has the same PA Risc processor (1.1) model as most of the
700s (except the new geckos and 715s).  Also, you should be aware
(if you aren't already) that there is a special pa-risc 1.1 version of
the math libraries that is only linked in if you add extra flags.  I
haven't checked how my copy of postgres 4.2 was linked, but I didn't
twiddle anything.

>< WARN:read on "myclass0": permission denied
>> REPLACE
>< WARN:read on "myclass0": permission denied
>> REPLACE
>< WARN:read on "myclass0": permission denied
>> DELETE
>< WARN:define rule on "myclass0": permission denied
>> DEFINE
>
>this is all permission checking stuff.  could the first "bmake"
>have been done by one user and then the test run by another?
>bmaking the script sets a user-name inside of it, and if the
>user who actually *runs* the script has the postgres superuser
>(create user) permission, this can cause the behavior shown.

I ran the original bmake as postgres and the regression bmake as
simmdan.  Should I have run the original bmake as simmdan and then
changed all the ownerships?  Or, should I somehow get bmake to change
the user-name in the regression script or something?

>re: timestamps
>i think what we'll have to do is set the $%@! timezone to GMT
>in the regression script.  it's sort of evil to do that because
>then you never really exercise the daylight savings time code,
>but it's probably better than wading through a hundred lines of
>meaningless diffs.  i flip through it quickly because i know
>what to look for, but ..

Sounds like that might be a good idea.  I will say, though, that I
noticed funny time stamps in the inversion file system by doing
"pls -D regression -lsga /" while the regression tests were running.
I don't know enough about it to know whether or not the times were
supposed to be this way, but I ran it a couple of times and found a
couple of files in each case they had dates with a year 1969...

>thanks for your help in working through this.  i admire your 
>patience (i tend to be short on that, as marc can attest.. :-).

No problem.  Anything I can do to help.

Danny

-------
Daniel Simmons          electronic mail : simmdan@isu.edu
Idaho State University       voice mail : (208) 236-3199
Computer Center              snail mail : Box 8037, Pocatello