[Stackless] alive+paused+unscheduled tasklet GC problem

Bob Ippolito bob at redivi.com
Wed Jan 28 10:06:04 CET 2004


On Jan 28, 2004, at 2:44 AM, Bob Ippolito wrote:

> It seems like GC of tasklets causes some funky stuff to happen in the 
> mainloop.
>
> I just checked in a new "old style" test to 
> 2.3/src/Stackless/test/test_gc.py
-- bad stuff --
> that breakpoint is where I noticed it.. it sets WHY_NOT, but is not 
> NULL!  The opcode is PRINT_NEWLINE, so what's happening is that it's 
> apparently running the next opcode before it throws the exception (it 
> does do that, after getting an XXX error).

Ok the bug (simply explained, after hours of stepping through GDB) was 
that it wasn't checking for an exception when it resumes from an 
interrupt.

There may be a more efficient way to check than PyErr_Occurred(), but 
I'm not so sure since I think the exception gets set before the stack 
switch?  I'm not totally familiar with this stuff yet, I've never done 
"goto" at this high of a level before ;)

Fixed in CVS.

wanting-a-real-issue-tracker-ly y'rs,

-bob

Oh, btw, it seems that we're not setting g+w wcvs by default when we 
add a new directory to CVS, which makes the anonymous users unhappy 
because they can't do a checkout (cvs daemon needs write access to 
create a lock file).  I fixed this for the unittests directory, because 
Donovan reported it to me and I figured it out, but I can't guarantee 
I'll remember next time :)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2357 bytes
Desc: not available
URL: <http://www.stackless.com/pipermail/stackless/attachments/20040128/edbf5be9/attachment.bin>
-------------- next part --------------
_______________________________________________
Stackless mailing list
Stackless at stackless.com
http://www.stackless.com/mailman/listinfo/stackless


More information about the Stackless mailing list