Open GoogleCodeExporter opened 9 years ago
I checked r527, the major change is store a global object reference in the JS
context object, it will ensure the global object will be free when JS context
destroyed.
btw: where is the 'method-wrapper' object? it seems cause the issue
Original comment by flier...@gmail.com
on 20 Aug 2013 at 3:26
The code which causes the TypeError could be read at
https://github.com/buffer/thug/blob/master/src/DOM/Window.py#L850
Please consider that I'm using V8 -r16223. With PyV8 < -r527 no problems at
all. The issue appears with -r527 (and successive releases).
Original comment by angelo.d...@gmail.com
on 20 Aug 2013 at 10:17
I did a simple test and seems like simply commenting out the line
Py_DECREF(global.ptr());
removes the issue. Hope it could be useful to you.
Original comment by angelo.d...@gmail.com
on 3 Sep 2013 at 7:19
After upgrading to V8 -r16670 and PyV8 -r536 I get the following error
"Fatal Python error: Inconsistent interned string state."
I tried again commenting out the line
Py_DECREF(global.ptr());
and this removes the issue as already pointed out in the previous comment.
Original comment by angelo.d...@gmail.com
on 13 Sep 2013 at 8:34
I was able to reproduce the issue and the fix (using thug to analyze
facebook.com)
I found the affected line in py_v8/src/Context.cpp.
Original comment by 0xC...@gmail.com
on 30 Sep 2013 at 9:59
Flier / Angelo,
"Fatal Python error: Inconsistent interned string state."
The patch looks fine but the above error happens when you call
JSEngine.collect(). Can you confirm? Do we still need to call
JSEngine.collect() with the patch in r527?
Thanks
Original comment by jmalo...@gmail.com
on 18 Nov 2013 at 4:54
Also, it seems like this only happens when you pass a Global object to
JSContext()?
Original comment by jmalo...@gmail.com
on 18 Nov 2013 at 6:23
To replicate - call JSEngine.collect() in
https://code.google.com/p/pyv8/source/browse/trunk/PyV8.py#1639 and that should
produce the error/python segfault.
Original comment by jmalo...@gmail.com
on 18 Nov 2013 at 6:36
OK, so disabling SUPPORT_TRACE_LIFECYCLE seems to fix the issue here.
Original comment by jmalo...@gmail.com
on 19 Nov 2013 at 12:52
Flier any update about this issue? Sorry to bother you but I would really like
to remove the need to patch PyV8 for Thug users. Thanks, Angelo.
Original comment by angelo.d...@gmail.com
on 5 Mar 2014 at 12:28
1 year after I suppose this issue it's still relevant because remains opened.
Any news about?
Original comment by rocknrol...@gmail.com
on 30 Mar 2015 at 10:58
Original issue reported on code.google.com by
angelo.d...@gmail.com
on 19 Aug 2013 at 5:57