We compare the result of PyRefcon against Pungi and RID literally through the comparison with CpyChecker in their papers 12. The data of Pungi is extracted from Table 2 on page 100 and Table 3 on page 101; whereas the data of RID is copied from Table 2 on page 542.
The number of true positives of PyRefcon, CpyChecker, and Pungi.
| Project | PyRefcon | CpyChecker | Pungi |
|---|---|---|---|
| pyaudio | 42 | 25 | 30 |
| pycrypto | 7 | 6 | 7 |
| pyxattr | 2 | 2 | 2 |
| rrdtool | 24 | 0 | 0 |
| dbus | 9 | 1 | 1 |
| duplicity | 3 | 2 | 2 |
| Total | 87 | 36 | 42 |
The number of true positives of PyRefcon, CpyChecker, and RID.
| Project | PyRefcon | CpyChecker | RID |
|---|---|---|---|
| pyaudio | 42 | 32 | 46 |
The corresponding results above is shown in Figure 9b.
Li, Siliang, and Gang Tan. “Finding reference-counting errors in Python/C programs with affine analysis.” In ECOOP 2014–Object-Oriented Programming: 28th European Conference, Uppsala, Sweden, July 28–August 1, 2014. Proceedings 28, pp. 80-104. Springer Berlin Heidelberg, 2014.↩
Mao, Junjie, Yu Chen, Qixue Xiao, and Yuanchun Shi. “RID: finding reference count bugs with inconsistent path pair checking.” In Proceedings of the Twenty-First International Conference on Architectural Support for Programming Languages and Operating Systems, pp. 531-544. 2016.↩