Perceived Crash (was: Stability of Sqeuak)

robin robin at sublime.org
Wed Aug 15 13:00:55 UTC 2001


Hi,

Ok - I've investigated my first perceived crash, and here's a lot more detail.  Andrew: sorry again for having frustrated you.

The platform is Windows 2000, fully service packed and patched.  System RAM is 192M.  CPU is a K6-2 500.

To reproduce:

1. Take a fresh Squeak 3.0 stable. ( the VM Reports itself as: 3.1 VM (alpha build 3) from Mar 15 2001).  Is this stable? I downloaded it from: 
http://isgwww.cs.uni-magdeburg.de/~raab/squeak/Squeak.zip
Following the link at: http://minnow.cc.gatech.edu/squeak/308

2. Create a new morphic project, and open it.

3. Open a workspace.

4. Enter the following and doIt:

a _ TextMorph new.
a openInWorld.
b _ Set new.
b add: a.


5. Then enter the following and printIt:

b storeString

At this point, I perceive the 'squeak system' to crash.  Clearly the vm isn't altogether dead as it will redraw its bitmap when windows asks it to.  However, I haven't been able to get it to respond. Memory usage of the squeak process starts to grow (it's at 140M and climbing as I
write) then at some point around 160M, it abruptly drops back down to around 14M and then starts growing again. Because of all the swapping, the whole platform is seriously compromised.

I've tried "ctrl-alt-." patiently several times to no avail. Is this the correct sequence?  

Closing the window doesn't bring up the "close without saving" dialogue - instead the windows "Program is not responding, end anyway" comes up.

Is there a way I could have single stepped into this to investigate myself?  My speculation is that the graph of object references is very large, or circular (and this is unhandled?). But that's just guesswork.

I hope that gives people enough to go on if they're interested.

-Robin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20010815/f9e7bf4c/attachment.htm


More information about the Squeak-dev mailing list