URGENT WARNING -- Don't snapshot! (was: "What is this error?")

Scott Wallace scott.wallace at squeakland.org
Mon Jul 1 03:38:47 UTC 2002


At 7:13 PM -0700 6/30/02, Tim Rowledge wrote:
>Scott Wallace <scott.wallace at squeakland.org> is claimed by the 
>authorities to have written:
>
>>  It appears that the latest official 3.3alpha and 3.2gamma images both
>>  have a significant, and perniciously dangerous, bug, which seemingly
>>  arrived in update 4897.
>That's curious - I've saved and restarted 4899 images a bunch of times
>on both Acorn and Mac/osx. What is the symptom?


The symptom is a VM error alert (in a Mac modal dialog if on a Mac) 
saying "Scheduler could not find a runnable process," appearing upon 
coming back from a snapshot -- either immediately after a "save" is 
issued in a running image or, after a save-and-quit, upon the next 
startup.  In either situation, the vm is dead in the water.

Behind the VM alert, on the Squeak desktop, a pre-debug notifier is 
visible whose message is "Error: attempt to evaluate a block that is 
already being evaluated".

It does not happen every time -- it is a timing-related issue.  But 
-- on the Mac anyway -- if you keep on making snapshots it seemingly 
will always *eventually* happen.  Sometimes it happens on the first 
or second save, sometimes it takes more.  And when it does happen, 
your image is hosed.  It's akin to Russian roulette.

The problem has been seen at least on the PC and on Mac/Classic.

   -- Scott


PS:  I wonder how you ever got an image with update level 4899?  The 
next burst of updates after 4889 took it straight to 4904 iirc.



More information about the Squeak-dev mailing list