Re: "Rude" behavior by developers

New Message Reply Date view Thread view Subject view Author view

From: Masatsugu Nagata (nagata@KURIMS.KYOTO-U.AC.JP)
Date: 12 Oct 98, 17:04 EST


From: Masatsugu Nagata <nagata@KURIMS.KYOTO-U.AC.JP>
Subject: Re: "Rude" behavior by developers

Hey Rowland,

I don't want to barge into the quarrel either (and I doubt I could,
considering that English is not my mother language), but here is just one
suggestion of mine, to you.

You wrote:

> despite the fact that none of the serious problems I've
> raised have been fixed.
>
Regarding Alpha's out-of-memory crashes, you have mentioned two kinds
of problems.  I think you have "raised" one of them explicitly enough,
but I think you have not "raised" the other one, which is the more
important one, as a useful-enough form, yet.

(1)  The first one is that if tempMem is OFF and if you try to open a file
whose size is larger than the current freeMem, then Alpha unexpectedly quits
without warning, forcing you to lose unsaved works.  This problem has been
quite explicitly stated by you, and should be a quite straightforward
problem to be taken care of, and, in my _personal_ guess/hope/expectation/
hunch/whatever, I _think_ it would probably be successfully found and
resolved _when_ Alpha's core application code is taken care of, in some
future, by Vince and Jonathan and a few other people.  Yes, I would think
there would be a reasonable chance.

(2)  But the other, more delicate crash, with tempMem ON, is not "raised"
by you, explicitly enough, in my opinion.  You wrote:

> Date:         Tue, 22 Sep 1998 03:23:42 +0100
> From: Rebecca and Rowland <rebecca@ASTRID.U-NET.COM>
> Subject:      Re: 7.1 keeps crashing
>
> At the time of the crash, I had quite a lot of files open: well over a
> dozen; none of the files was large (largest 52K).  I *think* I was
> switching from one window to another, and Alpha `just stopped' halfway
> through the job.  The mouse pointer moved, but mouse clicks did nothing
> (I waited for a minute before trying the mouse).

I have experienced quite a lot of beta-testing software.  In my experience,
only reporting  "I was editing some files and the application crashed...
it repeats somewhat often when many files are open" usually isn't helpful
at all, and doesn't usually result in the problem getting fixed.
(EVEN if the developer replies with a "I see, I'll look at it" -type mail.)

For a programmer (I am speaking in general) to be able to fix a particular
problem, he absolutely needs to reproduce it on his machine.  (Generally
speaking, that is.  There are, of course, exceptions.)


Jonathan wrote:

> [**]Roland has had recent problems with Alpha crashes and he as received a
> total of at least thirteen responses from Me, Vince, Moritz Borgmann,
> Philip Aker, and especially Mark Nagata. He's been getting extensive
> troubleshooting advice, which is the _most_ important part of fixing the
> problem. Until it's clear exactly what the nature of the problem is, it
> can't possibly be solved.
>
I interpret this last sentence of his to mean that the second problem,
that is, your intermittent crashes with tempMem ON, has not been "raised
explicitly enough.   (I, for one, cannot reproduce any such crash on my
68030 machine, although I don't edit multi-megabyte files -- 1200K files
would be my maximum)

Yes, you reported you've experienced some such problems, and another person
supported your experience by commenting that he has observed Alpha is more
stable when tempMem is OFF.  But the problem itself has not _yet_ been
made explicit enough.


So, Rowland, there are several ways:  You could just give up right now and
revert back to 6.52.  You are perfectly entitled to do so.  Or, you could
keep being patient and continue exploring Alpha's memory management
(with a non-programmer's eye, of course) with a hope that someday you could
stumble upon a more reproducible way of crashing Alpha's tempMem.

You also wrote:

> What I'm going to do is see how Alpha works with 4500K and temp mem.  If
> it seems okay, I'll try reducing Alpha's RAM allocation to 4000K and make
> regular checks on freeMem.
>
I think that's a great way to choose from.  If you are lucky and find more
explicit information and if you'll be able to report it to the ALPHA-D list,
then, everybody on the list, _especially_ those who are more "responsible",
would be really happy.

Good luck!

Ciao,

Mark
--
Mark Nagata
mailto:nagata@kurims.kyoto-u.ac.jp


New Message Reply Date view Thread view Subject view Author view

This archive was generated by das@users.sourceforge.net with hypermail 2b29 on 31 Mar 01, 11:50 EST