weird crash

[GoogleCodeExporter]

What steps will reproduce the problem?
1. i'd attached the example code
2.
3.

What is the expected output? What do you see instead?
Print from 0-999 and nothing. Instead i get a crash after 999.

What version of the product are you using? On what operating system?
PyV8 0.8 (win32-2.6 with python 2.6-win32 on x64 machine)

Please provide any additional information below.

Original issue reported on code.google.com by Liubomir...@gmail.com on 22 Feb 2010 at 2:59

Attachments:

• test_pyv8.py

[GoogleCodeExporter]

The same code, with little changes:
jsCtx.eval(str(
"""
var arr = new Array();
for(var i=0;i<1000;i++) {
    arr[i] = "a" + 1;
}
"""
))

arr = jsCtx.locals.arr
for k in arr:
    print k


Prints "stdout" ??

Original comment by Liubomir...@gmail.com on 22 Feb 2010 at 3:07

[GoogleCodeExporter]

I need to setup a x64 environment and reproduce it, could you reproduce it on 
x86 
system? I tried the latest SVN build, it seems ok

Original comment by flier...@gmail.com on 22 Feb 2010 at 6:00

• Changed state: Accepted

[GoogleCodeExporter]

Hm, my test machine is x64, but the installation files (python2.6 + pyv8 are x86
installation), so i dont think this is because of the x64 machine.
Currently i dont have proper Windows machine with x86 :(

I can test it on x86 (ubuntu) the next week if this will help ?

Original comment by Liubomir...@gmail.com on 23 Feb 2010 at 11:59

[GoogleCodeExporter]

ok, I will install a test machine to verify it soon, thanks anyway

Original comment by flier...@gmail.com on 23 Feb 2010 at 12:09

[GoogleCodeExporter]

10x for the good support :)
If you need something else, message me here, so i can help :)

Original comment by Liubomir...@gmail.com on 23 Feb 2010 at 12:13

[GoogleCodeExporter]

I have installed a x64 and x86 test machine, it seems PyV8 v0.8 definitely have 
issue 
to handle a very large array. I guess the root cause is the wrong reference 
count, 
because I have fixed several related issues in the latest SVN build, and it 
seems ok 
to run your test script.

So, could you verify it with the private build in attachment? I think I should 
make 
some paper work to release a new v0.9 version in this week, I wish it could 
solve 
your issue.

Thanks

Original comment by flier...@gmail.com on 24 Feb 2010 at 5:15

• Changed state: Fixed
• Added labels: OpSys-All

Attachments:

• PyV8-0.9.win32-py2.6.exe

[GoogleCodeExporter]

tested and it works :)
10x

Original comment by Liubomir...@gmail.com on 25 Feb 2010 at 6:43

[GoogleCodeExporter]

Thanks :)

Original comment by flier...@gmail.com on 25 Feb 2010 at 6:45

• Changed state: Verified

[GoogleCodeExporter]

Just wondering, is this bug valid in the linux pkg too ?
I'm about to push the new pyv8 on our productions, so just wondering to wait 
for the
new build or its already working (under linux x64) ?

Original comment by Liubomir...@gmail.com on 25 Feb 2010 at 6:54

[GoogleCodeExporter]

I think it should be a general issue, because the reference count doesn't 
depend on 
the platform. As you known, boost::python handle most of the reference count 
issues, 
but some special situation need be handled by manual.

In fact, the basic features has stabled a few weeks ago, but I had planned to 
release 
PyV8 v0.9 with the full feature Javascript AST support, but the plan was 
delayed :( 
So, the v0.9 will be released in this week without AST by default, because it 
is a 
very independent module.

It means, the SVN code will not be modified except the AST and build related 
issues 
:)

Original comment by flier...@gmail.com on 25 Feb 2010 at 7:06