squeak locks up on windows / help me debug it...

I thought the Pharo base image (3.10 with a few fonts & Balloon3d added) running on Windows was looking pretty stable (compared to other images). But after an hour of editing code on a Vista machine --an hour since the last save—Squeak has abandoned me. It is unresponsive. Trying both control and alt in combination with '.', 'l' (lower case L), 'c', does nothing apparently useful.

 

The process stack reported by Vista shows it to be incurring page-faults like mad. [OK, that has subsided while I typed this email.] My machine has 4 GB of RAM, I am running two instances of internet-explorer, and I have turned off the Windows search-indexing service. I know Vista has a guess-the-next-program pre-loader that could be a little overzealous in displacing programs from RAM… but still… it seems like this 101MB program shouldn't generate any page faults when the other two leading memory pigs are using less than double that much virtual memory. Vista's Resource Monitor claims I am using only 43% of physical memory. The page-fault count for Squeak.exe doubled while I typed this message.

 

When I click the Squeak icon on the window frame of the running copy (upper left on Windows), I do see the menu for VM Preferences. It shows me the menu and allows me to invoke actions. It will allow me to toggle the "Show output console" on and off. The output console shows me changes to memory (as I have that option checked). I can use my mouse wheel to scroll this console. Control-'.' and Control-c (and lower case L) in this console do not change anything.

 

After mucking around in the console area, I am now getting "WARNING: event buffer overflow" just moving my mouse around in squeak. After turning the console off, the pointer is now invisible except when you click. It appears that the overall squeak window is repainting itself properly.

 

-- Help me debug it

 

I have an interest in building a commercial application in Squeak. The FAQs says squeak is stable. Given the active community of developers (including The Great Ones), I imagine that it is stable, and that I am just doing something dumb… repeatedly. However, for Squeak to be adopted more broadly, it can't go zombie on users, even when the user does something dumb. That's an absolute.

 

What is the longest uptime a squeak image has ever endured (while doing something useful)? Which version was that and on what platform?

 

So I am willing to work to help make squeak more robust, but as a complete newbie with it, I need much guidance. So right now, with this hung-up image, what should I do? Should I attach to it with gdb? And if so, then what? [If the internal process control is working, I won't be able to make heads or tales out of it from gdb, right?]  Or can I launch another squeak and send it a signal in some way [is it listening?]? What is most useful in this case?

 

[Is there a way in Windows/Vista to signal an app so as to force a core dump?]

 

Thanks in advance,

Cam

PS: I want to emphasize that I do not care about the code lost in this particular instance (it was tutorial code), instead I am looking for a stable free development environment… and I'm willing to help make one.

 

__________ Information from ESET Smart Security, version of virus signature database 3755 (20090109) __________

The message was checked by ESET Smart Security.

http://www.eset.com

_______________________________________________
Beginners mailing list
[hidden email]
http://lists.squeakfoundation.org/mailman/listinfo/beginners

Squeak the development environment does not protect a developer from doing something dangerous. It probably won't ever, but maybe a file based Smalltalk like GNU does.

Smalltalk become: nil

is but one of a many ways to destroy your image (not saying that's what you did, just something dangerous to do).

By your definition of stable, Squeak would protect you from trashing your image. That's not how images work. I've also used VisualWorks and IBM VisualAge and they were both super-stable for deployment but similarly easy to crash (well, not nearly as easy to crash as Squeak, but still...).

Squeak applications can be very stable. I've had swikis run for months and months in the 2.x and early 3.x days on WinNT (this was many moons ago). Frankly, NT was more likely to tank than Squeak. Those images weren't safe from me if I was doing something dangerous, but the users couldn't do much harm just using the wiki.

I don't run anything on Squeak right now that has a long uptime (since even my web apps are suspended when I close my laptop), but I suspend and resume all the time on my deployed images (not development images) and they just keep on running. These are in Seaside.

The good news is your code isn't lost. You can roll it back from a clean image from the .changes file. This is handy, but it is even better to use something like Monticello.

Images are great for jump starting (loading a distro with useful tools) and great for returning right where you left off, but they aren't a substitute for version control. Images are really hard to repair once they've gone bad. I certainly don't know how and I've been doing Smalltalk since 95.

On Fri, Jan 9, 2009 at 6:39 PM, Cameron Sanders <[hidden email]> wrote:

I thought the Pharo base image (3.10 with a few fonts & Balloon3d added) running on Windows was looking pretty stable (compared to other images). But after an hour of editing code on a Vista machine --an hour since the last save—Squeak has abandoned me. It is unresponsive. Trying both control and alt in combination with '.', 'l' (lower case L), 'c', does nothing apparently useful.

 

The process stack reported by Vista shows it to be incurring page-faults like mad. [OK, that has subsided while I typed this email.] My machine has 4 GB of RAM, I am running two instances of internet-explorer, and I have turned off the Windows search-indexing service. I know Vista has a guess-the-next-program pre-loader that could be a little overzealous in displacing programs from RAM… but still… it seems like this 101MB program shouldn't generate any page faults when the other two leading memory pigs are using less than double that much virtual memory. Vista's Resource Monitor claims I am using only 43% of physical memory. The page-fault count for Squeak.exe doubled while I typed this message.

 

When I click the Squeak icon on the window frame of the running copy (upper left on Windows), I do see the menu for VM Preferences. It shows me the menu and allows me to invoke actions. It will allow me to toggle the "Show output console" on and off. The output console shows me changes to memory (as I have that option checked). I can use my mouse wheel to scroll this console. Control-'.' and Control-c (and lower case L) in this console do not change anything.

 

After mucking around in the console area, I am now getting "WARNING: event buffer overflow" just moving my mouse around in squeak. After turning the console off, the pointer is now invisible except when you click. It appears that the overall squeak window is repainting itself properly.

 

-- Help me debug it

 

I have an interest in building a commercial application in Squeak. The FAQs says squeak is stable. Given the active community of developers (including The Great Ones), I imagine that it is stable, and that I am just doing something dumb… repeatedly. However, for Squeak to be adopted more broadly, it can't go zombie on users, even when the user does something dumb. That's an absolute.

 

What is the longest uptime a squeak image has ever endured (while doing something useful)? Which version was that and on what platform?

 

So I am willing to work to help make squeak more robust, but as a complete newbie with it, I need much guidance. So right now, with this hung-up image, what should I do? Should I attach to it with gdb? And if so, then what? [If the internal process control is working, I won't be able to make heads or tales out of it from gdb, right?]  Or can I launch another squeak and send it a signal in some way [is it listening?]? What is most useful in this case?

 

[Is there a way in Windows/Vista to signal an app so as to force a core dump?]

 

Thanks in advance,

Cam

PS: I want to emphasize that I do not care about the code lost in this particular instance (it was tutorial code), instead I am looking for a stable free development environment… and I'm willing to help make one.

 

__________ Information from ESET Smart Security, version of virus signature database 3755 (20090109) __________

The message was checked by ESET Smart Security.

http://www.eset.com

_______________________________________________
Beginners mailing list
[hidden email]
http://lists.squeakfoundation.org/mailman/listinfo/beginners

I thought the Pharo base image (3.10 with a few fonts & Balloon3d added) running on Windows was looking pretty stable (compared to other images). But after an hour of editing code on a Vista machine --an hour since the last save—Squeak has abandoned me. It is unresponsive. Trying both control and alt in combination with '.', 'l' (lower case L), 'c', does nothing apparently useful.

 

The process stack reported by Vista shows it to be incurring page-faults like mad. [OK, that has subsided while I typed this email.] My machine has 4 GB of RAM, I am running two instances of internet-explorer, and I have turned off the Windows search-indexing service. I know Vista has a guess-the-next-program pre-loader that could be a little overzealous in displacing programs from RAM… but still… it seems like this 101MB program shouldn't generate any page faults when the other two leading memory pigs are using less than double that much virtual memory. Vista's Resource Monitor claims I am using only 43% of physical memory. The page-fault count for Squeak.exe doubled while I typed this message.

 

When I click the Squeak icon on the window frame of the running copy (upper left on Windows), I do see the menu for VM Preferences. It shows me the menu and allows me to invoke actions. It will allow me to toggle the "Show output console" on and off. The output console shows me changes to memory (as I have that option checked). I can use my mouse wheel to scroll this console. Control-'.' and Control-c (and lower case L) in this console do not change anything.

 

After mucking around in the console area, I am now getting "WARNING: event buffer overflow" just moving my mouse around in squeak. After turning the console off, the pointer is now invisible except when you click. It appears that the overall squeak window is repainting itself properly.

 

-- Help me debug it

 

I have an interest in building a commercial application in Squeak. The FAQs says squeak is stable. Given the active community of developers (including The Great Ones), I imagine that it is stable, and that I am just doing something dumb… repeatedly. However, for Squeak to be adopted more broadly, it can't go zombie on users, even when the user does something dumb. That's an absolute.

 

What is the longest uptime a squeak image has ever endured (while doing something useful)? Which version was that and on what platform?

 

So I am willing to work to help make squeak more robust, but as a complete newbie with it, I need much guidance. So right now, with this hung-up image, what should I do? Should I attach to it with gdb? And if so, then what? [If the internal process control is working, I won't be able to make heads or tales out of it from gdb, right?]  Or can I launch another squeak and send it a signal in some way [is it listening?]? What is most useful in this case?

 

[Is there a way in Windows/Vista to signal an app so as to force a core dump?]

 

Thanks in advance,

Cam

PS: I want to emphasize that I do not care about the code lost in this particular instance (it was tutorial code), instead I am looking for a stable free development environment… and I'm willing to help make one.

 

__________ Information from ESET Smart Security, version of virus signature database 3755 (20090109) __________

The message was checked by ESET Smart Security.

http://www.eset.com

_______________________________________________
Beginners mailing list
[hidden email]
http://lists.squeakfoundation.org/mailman/listinfo/beginners

Herbert Koenig writes:
HK> this is getting real crazy now.

Yep. This loss of the root world menu (and in some cases, the loss of all
interaction with objects on the world) makes the system unusable. Hopefully
I will get these issues figured out soon... because I really need to get
back to actual production coding. One month ago I was hopeful that Squeak
would be the platform I use for the next few years... but unless I have some
miracle break-through in the next few days, I will have to move on. [Sadly,
after so many problems, I am hesitant to do any prototyping in squeak... for
fear it will be lost time.]

By the way, I wrote:
CS> world/morph window is not popping up.

But I should have said: world/morph "menu" is not popping up.

--
HK> In case it doesn't, alt K should open a workspace there you type:
World putUpWorldMenuFromEscapeKey
and doIt.

Thank you, that helps me get oriented, and gives me something concrete to
try. In this (last) case where the mouse buttons would not raise a menu, the
escape key did. However, the menu just disappeared without effect when I
clicked it, and no mouse-over visual effects before clicking.

Next time I will try to raise a menu via code, as you have shown me.
--

HK> Maybe you need to set your display depth to Window's display depth.

World menu, appearence, set display depth.

I will check this particular image, but so far these have all been set to 32
bits.

> And you should look at your virus scanner.

Hmmm... an interesting thought. Is there some known bug?

CS> Any tips for the easiest way to re-attach the world menus, should I ever
be
CS> in a position to shout?

HK> My Denglisch (german english) can't parse that :-))

You already covered how to bring up the world menu via code. Here I am
asking how to re-bind the base menus (on mouse click events)? (Assuming the
menus exist in code -- I understand how things evolve in smalltalk.) I will
peruse the code more.

A version of the workspace is called Shout (as I understand it)... and so I
simply meant that "if I can bring up a workspace".

Cheers is right. Thanks for some direction!

I attached gdb (gnu debugger) to the process of squeak in the case where the
world was not responding to mouse or keyboard inputs (even though the
window-frame menu could effect the console), but I did not know what
functions to watch for. If someone could have told me to "see if
XxxEventHandlerPassthroughYyyy() is ever called", or something, I could have
done it. As it was, I would "step 1000", for example, check the stack, and
very often it was in a GC function, although not always.

Should I generate the VM code and get oriented with it, or is the outer-most
morphic world (instance) created through a dll or something that does not
require specific code in the VM? Also, I'm wondering what signals it will
respond to -- perhaps I can add a few back-door debugger launchers.

Ugh... so much to learn, and so little time.

Thanks again!
Cam

_______________________________________________
Beginners mailing list
[hidden email]
http://lists.squeakfoundation.org/mailman/listinfo/beginners