Squeak hangs when AVG virus scan is run

As the subject says: Squeak hangs if an AVG anti-virus scan is run. I
know, because I have one scheduled for 8 am on Saturdays and exactly at
8 Morphic locked up solid and stopped processing events.

Contrast with the email client I'm using, which is a touch sluggish but
otherwise working perfectly.

There's been no response that I could detect (even scanning the archives
again) to my response to the VM crash thread.

There are also performance issues. Performing 100,000 math operations on
floats and small integers with some message passing, looping, and block
calling takes a full second or two on a 2GHz machine. Is that normal?

John Ersatznom wrote:
> As the subject says: Squeak hangs if an AVG anti-virus scan is run. I
> know, because I have one scheduled for 8 am on Saturdays and exactly at
> 8 Morphic locked up solid and stopped processing events.

It's worse. The virus scan completing did not result in Squeak
recovering. It still keeps flashing "Warning: event buffer overflow" and
otherwise acting hung. That fresh mouse events aren't finding room to be
appended to the buffer means that the buffer isn't being emptied from
the other end, even though the original cause of the problem is gone.
Apparently it isn't even trying to catch up on the queued-up events and
begin responding again, and it looks like I'll have to resort once more
to the three-fingered salute.

I'm getting rather tired of Squeak crashing at the drop of a hat. My
Java development tools do not do so. My C/C++ development tools do not
do so. My other applications do not do so -- not even the ones from
Microsoft, these days. Those go or wonk out on a weekly basis but not a
daily one.

Is there a tweak to make it much more robust, or is it just a matter of
giving it time to mature more?

On Sat, 13 Jan 2007 14:07:51 +0100, John Ersatznom wrote:

...
> There are also performance issues. Performing 100,000 math operations on  
> floats and small integers with some message passing, looping, and block  
> calling takes a full second or two on a 2GHz machine. Is that normal?

Sure, that's quite normal ;-) For example

  [(1 to: 123456) collect: [:smi | 1.0 raisedTo: smi]] timeToRun

takes 2 seconds here on my 1.73 GHz notebook ;-)

/Klaus

Hi John,

Check out: http://weeklysqueak.wordpress.com/2006/10/26/image-freezing/ 

What I've seen is that there are problems with switching processes of the
same priority.  The problem I saw was an application that was running on a
separate thread at the idle priority.  Once I bumped the process up by one
the freezing went away.  

I've also see some problems with threading as you mentioned.  The issue
appears to be that thread switching works fine if you have squeak as the
selected application.  Once you click off it, it appears to have some
trouble switching threads.  I wonder if your problem with AVG is not related
to what I saw.  I fixed my problem by managing the treads myself.  I used
the class Monitor and coded my processes by hand.  Monitor is very cool!

You could also try turning on CPUWatcher.  World menu>open>process
browser>start CPUWatcher.  This will suspend processes that are running away
with your app and ask you if you would like to debug at a lower priority.
That gives you the chance to figure out what's going wrong.

Now if you are not doing anything unusual with processes then you can
disregard all these comments.

Ron Teitelbaum
President / Principal Software Engineer
US Medical Record Specialists
[hidden email]

(Note: This is in reply to Ron's post. Unfortunately, if I'm not copied
as a separate recipient it doesn't seem I can really thread replies
properly, without subscribing to the list and inundating my inbox anyway.)

I had actually user-started a Process or three recently, but this seems
an unlikely culprit:

1. The hang coincided *exactly* with the start of the virus scan; I'd
    been using [foo] newProcess resume for a bit without problems other
    than with non-reentrancy of blocks.
2. Said processes actually being started with [foo] newProcess
    setPriority: Processor lowestPriority; resume anyway. (Or maybe just
    priority: or whatever; don't recall offhand.)
3. Task manager showed (and still shows) the hung Squeak process using
    0% CPU. A Process hogging the CPU away from the UI handler would
    saturate it instead. Instead the UI event handling thread seems to
    have been starved, or begun blocking until a condition that is never
    actually met. This smells like starvation or a deadlock to me, and it
    seems to involve the interaction with native multitasking. Perhaps
    I/O blocking for long enough risks provoking a latent
    deadlock-causing bug at the low levels of the Windows implementation;
    a virus-scan causes heavy contention for disk I/O.

John Ersatznom wrote:
If this is reproducible on other systems, perhaps you should file a bug.
Especially if it happens on a fresh image (i.e., not your code. This
would be a real problem for production sw in the field.

--
brad fuller
www.bradfuller.com

That's interesting. I also use AVG, and I have noticed the same problem (on
Windows 2000).
I hadn't thought that it could be AVG causing the problem. Now it makes sense.

Interestingly, it doesn't seem to happen when Squeak is minimized.

Sometimes I can delete the error messages from the console, then hide the
console window, and the image is useable, though slow.

Cheers,
Andy

----- Original Message -----
From: "John Ersatznom" <[hidden email]>
To: <[hidden email]>
Sent: Saturday, January 13, 2007 1:07 PM
Subject: Squeak hangs when AVG virus scan is run

Hm, so this is a Squeak issue or a AVG issue? Is this happening with
other Anti-Virus products?
I'm working at Grisoft so maybe could fix it if AVG does the problem ;-)

Elod

Elod Kironsky wrote:
> Hm, so this is a Squeak issue or a AVG issue?

I don't think we have enough data to point either way. I use AVG myself
and so far didn't have any problems.

> Is this happening with other Anti-Virus products?

Many moons back we had some problems with (I think) Norton Antivirus.
However, those went away after some update to their software (Squeak was
not the only software affected by it).

> I'm working at Grisoft so maybe could fix it if AVG does the problem ;-)

Well, if you have heard about similar problems caused by AVG it would
certainly be interesting to see if there is any correlation. If you have
any guidelines about particular behavior that might be problematic it
would be good to know about that, too.

Cheers,
   - Andreas

Hi John,

My question is if AVG running a scan has similar memory ramifications to
selecting another application (allowing squeak to run in background).  My
experience was that when I clicked off squeak the processes that were
switching properly stopped working efficiently, and I eventually ended up
with a frozen image and eventually the buffer overflow messages you
mentioned.

> 2. Said processes actually being started with [foo] newProcess
>     setPriority: Processor lowestPriority; resume anyway. (Or maybe just
>     priority: or whatever; don't recall offhand.)

There was a separate problem with a process running in lowestPriority.  The
loop caused a similar problem.  This problem was solved by raising the
priority.  I still believe there is a problem with process switching
especially with a thread running at the same priority as idle.  I do not
know if this is a 3.9 specific problem but a number of people did say that
they were experiencing the same problem and were not switching to 3.9
because of that.

> 3. Task manager showed (and still shows) the hung Squeak process using
>     0% CPU. A Process hogging the CPU away from the UI handler would
>     saturate it instead. Instead the UI event handling thread seems to
>     have been starved, or begun blocking until a condition that is never
>     actually met. This smells like starvation or a deadlock to me, and it
>     seems to involve the interaction with native multitasking. Perhaps
>     I/O blocking for long enough risks provoking a latent
>     deadlock-causing bug at the low levels of the Windows implementation;
>     a virus-scan causes heavy contention for disk I/O.

This is the same as what I experienced too.  I run on a laptop and when I'm
doing some heaver processing my fan let's me know loud and clear.  I know
that my image is frozen when the fan stops!  I am not sure exactly what
happens when the image freezes but it stops hogging resources, has no mouse
or keyboard responses, and I can usually get responses from the vm by
selecting debugging options, and it's odd to me that clicking the close box
on the window still brings up save before quit?  Also I can get the mouse to
click in the console window.

I would be great if we could figure this one out.

Ron Teitelbaum