🚀 go-pugleaf

jamesb@ecst.csuchico.edu (James L. Brookes) writes:

>So, is it worth adding?  Comments anyone?

  IMHO, no.  The ORCA method was created because at one time GS/OS
didn't support restarting through the Quit call.  It does now,
and thus is the standard.  If a program does rexit(), then it should
be restartable, period.  If the memory is needed the Memory Manager
will purge the restartable image anyway.

-- 
 Jawaid Bazyar              |   Ask me about the GNO Multitasking Environment
 Procyon, Inc.              |   for the Apple IIgs!   
 bazyar@cs.uiuc.edu         |   P.O Box 620334
 --Apple II Forever!--      |   Littleton, CO 80162-0334   (303) 933-4649

In article <bazyar.726209355@teal> bazyar@teal.csn.org (Jawaid Bazyar) writes:
>jamesb@ecst.csuchico.edu (James L. Brookes) writes:
>
>>So, is it worth adding?  Comments anyone?

One reason I asked this, is because uEmacs has a variable to control
restartibility.  Does anyone actually turn this off?

>  IMHO, no.  The ORCA method was created because at one time GS/OS
>didn't support restarting through the Quit call.  It does now,
>and thus is the standard.  If a program does rexit(), then it should
>be restartable, period.  If the memory is needed the Memory Manager
>will purge the restartable image anyway.

So, then should all apps be written to use rexit()?  I don't really see
any reason to -not- make an app restartable, but then I have 4.25 megs
of memory, and want -everything- restartable :)

Hm, a couple of questions I just thought of re: restartability:

I've noticed that, if I have a restartable program running in the
background, and try to run the same application in the foreground, it will
not restart.  No surprise, that's to be expected.  But, even after one
execution, this copy of the app won't restart.  So, is it to be presumed
that there can only be one copy of a given restartable application in memory
at a time?  Is it possible, in a multiuser system, for someone (for example)
on console to use a restartable program, then exit, and for another person
not on console to use the copy in memory?

On the i/o-slow GS, restartability would be a great boon in a multiuser
system.  Running a cached application would allow the rest of the system
to operate at the same time, instead of lagging as the disk is accessed.
Allowing multiple copies of cached applications in memory would be nice
but I guess that isn't the way it works.

> Jawaid Bazyar              |   Ask me about the GNO Multitasking Environment
> bazyar@cs.uiuc.edu         |   P.O Box 620334
-- 
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
|          James Brookes            | Inet: bb252@cleveland.freenet.edu |
|``What, me worry?'' - A. E. Neuman |   jamesb@cscihp.ecst.csuchico.edu |
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

jamesb@ecst.csuchico.edu (James L. Brookes) writes:

>One reason I asked this, is because uEmacs has a variable to control
>restartibility.  Does anyone actually turn this off?

  Why? :-)

>So, then should all apps be written to use rexit()?  I don't really see
>any reason to -not- make an app restartable, but then I have 4.25 megs
>of memory, and want -everything- restartable :)

  Well, some applications would take a LOT of work to make restartable.
That is, excepting the trick that Morgan Davis used to make uEmacs
restartable.  I don't know that his technique will work with ORCA/C,
though.

>Hm, a couple of questions I just thought of re: restartability:

>I've noticed that, if I have a restartable program running in the
>background, and try to run the same application in the foreground, it will
>not restart.  No surprise, that's to be expected.  But, even after one
>execution, this copy of the app won't restart.  So, is it to be presumed
>that there can only be one copy of a given restartable application in memory
>at a time?  Is it possible, in a multiuser system, for someone (for example)
>on console to use a restartable program, then exit, and for another person
>not on console to use the copy in memory?

  Yes, but not simultaneously.  GNO specifcally reloads a program from
disk if a restartable executable is running and another copy is
requested.  The system loader simply doesn't deal with more than
one restartable copy of a program at one time.

-- 
 Jawaid Bazyar              |   Ask me about the GNO Multitasking Environment
 Procyon, Inc.              |   for the Apple IIgs!   
 bazyar@cs.uiuc.edu         |   P.O Box 620334
 --Apple II Forever!--      |   Littleton, CO 80162-0334   (303) 933-4649

jamesb@ecst.csuchico.edu (James L. Brookes) writes:

[Stuff about adding SYSCMND to GNO]

>So, is it worth adding?  Comments anyone?

Buy why? If a program assigns constants to variables then it should exit() and
if it was written to be restartable it should rexit(). That's all there is to
it. There is no reason to let the user control the restartability.

To let the user change the restartability implies that some people might not
like the default. But...

If the program is restartable and you want to make it non restartable, well...
  Why would you? No gain...

If the program is non restartable and you want to make it restartable, well...
  Why would you? You'd screw up your system.

>|          James Brookes            | Inet: bb252@cleveland.freenet.edu |
 ______________________________________________________________________
|~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~~~~|
| Phillip Vandry | Vandry@Cam.Org | II the future! | II 1993!          |
|________________|________________|________________|___________________|
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

jamesb@ecst.csuchico.edu (James L. Brookes) writes:

>I've noticed that, if I have a restartable program running in the
>background, and try to run the same application in the foreground, it will
>not restart.  No surprise, that's to be expected.  But, even after one
>execution, this copy of the app won't restart.  So, is it to be presumed
>that there can only be one copy of a given restartable application in memory
>at a time?  Is it possible, in a multiuser system, for someone (for example)
>on console to use a restartable program, then exit, and for another person
>not on console to use the copy in memory?

Oh, wait a minute. You bring up some interesting things here.

Let's say you start a second instance of a restartable program. Is it logical
to simply duplicate the existing copy rather than get it from disk? I think
not for two reasons. 1. You can't relocate it again. 2. [Explained below]

Also, if a process is kill(SIGKILL)ed, (or other "destructive" signal), it 
should never be restarted from memory. Here's why:

If I wrote a program that changes itself (or even changes a GS/OS parameter
block) then changes itself back (which I do often) I would make this program
restartable (assuming the rest of the program is restartable). Sure it assumes
that a certain address will contain a certain value, and that value gets 
changed later, but it gets changed right back, and unless the program exits
between the changing and the changing back, it's ok. The problem is that the
SIGKILL could easily occur between the two.

Example:

main      lda refnum
          sta readrec+2
          ReadGS readrec
          lda #500
          sta len
          ReadGS readrec
          lda #200
          sta len
          ...
readrec   dc  i'2'
          ds  2
          dc  a4'buffer'
len       dc  i4'200'
          ds  4

This fragment reads 700 bytes. The first read assumes that len will be set to
200. Unless the program exits while it is temporarily at 500, it will in fact
be at 200.

Yesterday I would have made this program restartable, but now I see that it is
not since the program could die while len is set to 500.

What is the solution?

>|          James Brookes            | Inet: bb252@cleveland.freenet.edu |
 ______________________________________________________________________
|~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~~~~|
| Phillip Vandry | Vandry@Cam.Org | II the future! | II 1993!          |
|________________|________________|________________|___________________|
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

In article <1icpfiINNd21@charnel.ecst.csuchico.edu> jamesb@ecst.csuchico.edu (James L. Brookes) writes:
> 
> So, then should all apps be written to use rexit()?  I don't really see
> any reason to -not- make an app restartable, but then I have 4.25 megs
> of memory, and want -everything- restartable :)

If a program relies on its variables being cleared on startup, and
they've been modified by a previous execution, it wouldn't be
restartable.  Making the variable segment 'reload' can get around this
problem (but this isn't easy to do with the standard linker).

> Hm, a couple of questions I just thought of re: restartability:
> 
> I've noticed that, if I have a restartable program running in the
> background, and try to run the same application in the foreground, it will
> not restart.  No surprise, that's to be expected.  But, even after one
> execution, this copy of the app won't restart.  So, is it to be presumed
> that there can only be one copy of a given restartable application in memory
> at a time?

The System Loader keeps track of the application's pathname and user
id.  It is not safe to have the same user ID for two different
pathnames; nor is it safe to have two different user IDs for the same
pathname.

> Is it possible, in a multiuser system, for someone (for example)
> on console to use a restartable program, then exit, and for another person
> not on console to use the copy in memory?

For simultaneous use by two people, the program would require two sets
of variables.  Since the addresses of the variables are patched into
the program at load time, you would need to load a new copy of the
program to refer to a different set of variables.

It can't be done, with the current System Loader architecture.
-- 
David Empson

Internet: David.Empson@bbs.actrix.gen.nz    EMPSON_D@kosmos.wcc.govt.nz
Snail mail: P.O. Box 27-103, Wellington, New Zealand