[New-bugs-announce] [issue23836] PEP 475: Enhance faulthandler to handle better reentrant calls

[STINNER Victor]

New submission from STINNER Victor:

While working on the PEP 475, I noticed that the faulthandler module doesn't handle well reentrant calls to its signal handlers. If a signal is received while faulthandler is displaying the traceback, faulthandler restarts to display a new traceback (and then finish to display the previous traceback).

Moreover, faulthandler uses the C function write() to display the traceback and it ignores write() errors. It should retry write() if write() is interrupted by a signal (fails with EINTR).

Attached py_write_noraise.patch:

- Add a new _Py_write_noraise() function
- Document functions releasing the GIL

Attached traceback_eintr.patch:

- Modify functions displaying traceback to use _Py_write_noraise() instead of write()

Attached faulthandler_eintr.patch:

- Add a reentrant flag to the two signal handlers of faulthandler
- Modify faulthandler.dump_traceback() to call PyErr_CheckSignals(), to call the Python signal handler if a signal was received while faulthandler was displaying the traceback of all Python threads

----------
files: py_write_noraise.patch
keywords: patch
messages: 239776
nosy: haypo
priority: normal
severity: normal
status: open
title: PEP 475: Enhance faulthandler to handle better reentrant calls
versions: Python 3.4, Python 3.5
Added file: http://bugs.python.org/file38775/py_write_noraise.patch

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue23836>
_______________________________________