[pypy-dev] "import signal" fails when running PyPy interpreted

[Stefan Behnel]

Armin Rigo, 11.03.2012 20:50:
> Did you realize that you said in the *same* e-mail:
> 
>> You should put up a "beginners guide to debugging bugs in PyPy" somewhere.
> 
> and
> 
>>> It's tersely documented in pypy/doc/coding_guide.rst.
>> I'm not interested in details.
> 
> Sorry if this was not your intention but it sounds to me a lot like
> "you need to write docs, but I'm not going to read it anyway".

"details" is different from a "beginners guide to debugging", at least the
way I understand it. What I'd like to have is a short step by step guide
saying what I need to do, and how to set up a working environment, in order
to efficiently debug into a bug in PyPy, specifically its cpyext module. I
tried two ways so far (nightly build in gdb and interpreted source) and
they both proved to be dead ends.

Here's the trial and error guide for Cython, as an example of what I mean:

http://wiki.cython.org/HackerGuide#Gettingstarted

To be clear, I happily don't care how PyPy works internally and I don't
want to waste my time learning it. I just want to find out what's wrong in
its cpyext layer by analysing the bugs that currently block it from running
Cython generated code.

OTOH, if you prefer to debug and fix everything yourself, I'm fine with
that also. You can look at the guide above, it will tell you how to run the
test suite, and that will quickly show you the crashes, so you can debug
them. Here's a list of test failures:

https://sage.math.washington.edu:8091/hudson/view/dev-scoder/job/cython-scoder-pypy-nightly/lastCompletedBuild/testReport/

and here are the latest builds, the red ones are eventual crashes (using
jitted PyPy up to build 48):

https://sage.math.washington.edu:8091/hudson/view/dev-scoder/job/cython-scoder-pypy-nightly/

A couple of crashers are actually disabled in the PyPy runs, you'll notice
them when you run the tests yourself. The list is written into the build log.

Stefan