[pypy-issue] Issue #1926: PyPy STM segfault with short transactions (pypy/pypy)

Konstantin Lopuhin issues-reply at bitbucket.org
Fri Nov 14 07:56:14 CET 2014

New issue 1926: PyPy STM segfault with short transactions

Konstantin Lopuhin:

Steps to reproduce are the same as in #1918:

hg clone ssh://hg@bitbucket.org/kostialopuhin/tornado-stm-bench
cd tornado-stm-bench
hg up a038bf99de71
virtualenv env -p pypy-c-r74378-74379-stm-jit
source env/bin/activate
pip install -r requirements.txt
./primes.py 4
the only difference is that now to run benchmark do

siege -c 20 -t 20s http://localhost:8888/500

the difference is that here transactions are much shorter (primes up to 500 instead of 10000). This is not 100% reliable way to get a segfault, but I get it more often than not.

I can try with a newer build

More information about the pypy-issue mailing list