[reportlab-users] Crash on debug builds of python 3.4.3

[Robin Becker]

I haven't run the tests on anything other than official builds. I guess I should 
try and build a pydebug version of 3.4.x with linux as that is much much easier 
than getting VS xxxx to run properly. I have one build machine for windows and 
that is running builds for 2.7 3.3, 3.4 & 3.5 and I really don't want to mess it 
up. It sounds as though you could have some kind of issue with one or more C 
extensions. For the RL toolkit you should be able to remove the rl_accel pyd and 
see if that improves things.

If that fixes stuff then at least we know where the problem lies. If it's in the 
python code all bets are off.
-- 
Robin Becker
On 11/01/2016 17:41, Mark De Wit wrote:
> Hi,
>
> I'm working on embedding Python into my application as a scripting tool.  I have therefore recompiled Python 3.4.3 using Visual Studio 2013, in both debug and release versions.  I have also built the ReportLab 3.2.0 source release from pypi using the same environment.
>
> When running the simplest of ReportLab samples (copy/paste from the user guide), I find that my Python interpreter crashes on shutdown (specifically, when calling Py_Finalize).  The error appears to be a negative reference count on an object in the main Dict.  The release version appears to work ok, though whether this is due to less error checking I don't know... Ignorance is perhaps bliss in this scenario.  Note that this happens both inside my application (embedded interpreter) and the official python shell (python_d.exe crashed at the end of generating the user guide).
>
> I guess my question is - is ReportLabs tested against debug builds of Python?  Is the error genuine or false-positive (it's not too benign, the Python interpreter calls abort() for this fatal error...)?
>
> Kind regards,
> Mark de Wit
.....