What a day. I loaded 5.0 PUSH on the development system yesterday and
have spent every waking moment of today responding to issues. Allow me
to waste some bandwidth on insights and questions.
First, the library problem I mentioned eariler was bogus. We encountered
two problems that together looked like an library issues, but were not.
One entails running a program across account bounderies with ;DEBUG
(it's not working for us at the moment). I have called this into PICS but
they and I have linked up yet.
The other problem is amusing. Well, entertaining. OK. I needed a few
drinks before I found any humor in it. Anyway, a master process here
was failing, and falsely reporting file errors during CreateProcess.
Once we found the suspect code, and removed the Swahili error message,
and inserted another (which someone years from now will swear is written in
Sanskrit) we discovered a bad parm being passed to CreateProcess.
Funny. The code worked the day before, bad parm and all. Well,
seems some long forgotten programmer had assembled his itemnum/item array
with all available numbers from one to twenty-something. In there is
itemnum 13, which the intrinsic manual said was not used. Our code passed
an item of zero along with the useless itemnum 13, which in all previous
versions of MPE/iX back to 1.2 caused no problem.
Now it does. I won't beat on HP. They said -- more or less -- that we
shouldn't use 13, but we thought they were kidding.
While scrambling on this problem, we discovered 4.0 compiler .vs. LINKEDIT
problems. Since our code is a little complex -- in much the same way that
Hitler was a little rude -- we tend to copy old compilers into place when
installing the new version of the OS (i.e., run 4.0 compilers on a 5.0
system). This has always worked, though HP has never been comfortable
our doing so and refuses to support us. I can deal with that as long as
our programmers are cutting 4.0 code for the 4.0 production machines.
But in all this time, LINKEDIT never caused a problem. That is a new
version of LINKEDIT with the old compilers. Given the massive changes in
libraries in 5.0, I should have assumed LINKEDIT would behave *very*
differently -- and boy did it. We took the suspect code mentioned above,
compiled, linked without error (changing only the bogus itemnum 13
to a different value) and the program aborted at DBOpen with a
mangled database name and an open mode of 3462.
Yeah. I said the same thing, but with less polite language. We debated
on the best bet, and decided to copy the old LINKEDIT into PUB.SYS.
It and the compilers *appear* to be playing nicely now, but I'm not
cormfortable yet. Has anyone else tried this, and what have been your
long term results?
So now everything was working well, but there was this angry mob of
programmers outside who were chanting something about "Death to Guy for
mucking with VI", or some such. When I loaded 5.0, I renamed the old
CSL VI program (Stevie) back and let programmers use VI.HPBIN.SYS.
That was a mistake. You see, we have a boat load of HP UNIX boxes here,
and I foolishly expected the VI on MPE/iX to work almost like the VI on
HP-UX. Of course I still believe in the Easter Bunny an Santa. Not only
does it not work as well, or even vaguely like the HP-UX version, it
doesn't work as well as Stevie. Go figure. [Page/Up] and [Page/Dn]
do little more than beep and occationaly deficate on your screen. When
I smelled the tar and poultry waifting in from the hall, I decided to
tell everyone where I hid Stevie.
They haven't called about grep and awk yet, so maybe I'm safe -- for today.
Now, a question. I'm ignorant on how mail really works under the sheets.
I did notice that mailx was included in 5.0, and that I could send mail
to myself and others on the system. Messages don't leave the system, and
I don't see the sendmail daemon that I do on UNIX. Was there some issue
in porting sendmail, or is sendmail simply not in the POSIX 1/3 specs?
Or is HP being kind to the 3rd party vendors who make mail packages?
Does anyone know of a PD sendmail, or is anyone hacking a version as
we speak? Inquiring minds, and all that rot.
Speaking of 3rd party software, we had two packages go belly up. Both
checked version numbers, both knew how to handle X.50.20 (pull) but
not C.50.00 (push) (seems we got ahead of ourselves and ordered new
tapes from vendors before push was released and they were not sure what
the version numbers might be).
I'd like to take the opportunity to ask HP to . . . . . .
GET A CLUE ABOUT VERSION NUMBERS!
Sorry for the shouting, but this has transcended silly and plunged off
the cliffs of absurd. 5.0 divided between PULL and PUSH is weird. I
understand come June, instead of 5.1 (which is what PUSH should have
been called) we will have a 5.0 Express Release, which will have version
number 4.Q.2. Heaven help me for ever saying something nice about DEC,
but when I managed VMS systems, they never repeated or skipped a
version sequence except at major release time. This level of inanity
can only be the work of marketing managers, lawyers, or the two in
collusion.
Well, thanks for letting me waste some of your time. I feel much better
now, and will be feeling real good when the medication kicks in.
=======================================================================
Guy Smith Voice: 804-527-4000 ext 6664
Circuit City Stores, Inc. FAX: 804-527-4008
9950 Mayland Drive E-Mail: [log in to unmask]
Richmond, VA 23233-1464 Private E-Mail: [log in to unmask]
The thoughts expressed herein are mine and do not reflect those of my
employer, or anyone with common sense.
|
