Closed vstinner closed 1 year ago
In Python 3.6, there was a WANT_SIGFPE_HANDLER macro defined if Python was built with ./configure --with-fpectl. By default, the macro was not defined. The fpectl module installed a SIGFPE signal handler which uses longjmp(). If the WANT_SIGFPE_HANDLER macro was defined, PyFPE_START_PROTECT() and PyFPE_END_PROTECT() macros used setjmp() to catch SIGFPE.
All of this machinery was removed by bpo-29137 to fix ABI issues:
commit 735ae8d139a673b30b321dc10acfd3d14f0d633b Author: Nathaniel J. Smith \njs@pobox.com\ Date: Fri Jan 5 23:15:34 2018 -0800
bpo-29137: Remove fpectl module (bpo-4789)
This module has never been enabled by default, never worked correctly
on x86-64, and caused ABI problems that caused C extension
compatibility. See bpo-29137 for details/discussion.
The PyFPE_START_PROTECT() and PyFPE_END_PROTECT() macros still exist for backward compatibility with Python 3.6 and older. Python/pyfpe.c contains the comment:
/* These variables used to be used when Python was built with --with-fpectl,
I'm not convinced that these macros are part of the stable ABI. The PEP-384 doesn't exclude pyfpe.h explicitly, but it's also unclear if it's part of the stable ABI or not. "PyFPE_START_PROTECT" and "PyFPE_END_PROTECT" symbols are not part of the stable ABI at least: they are missing from PC/python3dll.c and Doc/data/stable_abi.dat. Well, these are macros which emitted directly code rather than calling an opaque function. That sounds really bad in term of stable ABI :-(
Python 3.6 macros, when WANT_SIGFPE_HANDLER macro is defined:
---
#include <signal.h>
#include <setjmp.h>
#include <math.h>
extern jmp_buf PyFPE_jbuf;
extern int PyFPE_counter;
extern double PyFPE_dummy(void *);
#define PyFPE_START_PROTECT(err_string, leave_stmt) \
if (!PyFPE_counter++ && setjmp(PyFPE_jbuf)) { \
PyErr_SetString(PyExc_FloatingPointError, err_string); \
PyFPE_counter = 0; \
leave_stmt; \
}
#define PyFPE_END_PROTECT(v) PyFPE_counter -= (int)PyFPE_dummy(&(v));
I propose to either deprecate these macros or remove them. These macros are not documented.
In Python 3.6, the PyFPE macros were used in the Python stdlib, by files:
Doc/library/fpectl.rst Include/pyfpe.h Modules/_tkinter.c Modules/clinic/cmathmodule.c.h Modules/cmathmodule.c Modules/fpetestmodule.c Modules/mathmodule.c Objects/complexobject.c Objects/floatobject.c Python/bltinmodule.c Python/pystrtod.c
But it doesn't seem to be used outside Python.
Note: In master, they are not used anywhere, they are only defined by Include/pyfpe.h.
Latest commit related to these macros:
commit be143ec99674ba38c5811f34cdb85ef39c2dc8f8 Author: Victor Stinner \vstinner@python.org\ Date: Wed Nov 20 02:51:30 2019 +0100
bpo-38835: Don't use PyFPE_START_PROTECT and PyFPE_END_PROTECT (GH-17231)
The PyFPE_START_PROTECT() and PyFPE_END_PROTECT() macros are empty:
they have been doing nothing for the last year (since commit
735ae8d139a673b30b321dc10acfd3d14f0d633b), so stop using them.
Another related change from bpo-37474:
commit 5e0ea7540f577c9684e272000fdfc80d29bb78a2 Author: Victor Stinner \vstinner@python.org\ Date: Tue Oct 1 13:12:29 2019 +0200
bpo-37474: Don't call fedisableexcept() on FreeBSD (GH-16515)
On FreeBSD, Python no longer calls fedisableexcept() at startup to
control the floating point control mode. The call became useless
since FreeBSD 6: it became the default mode.
I explicitly excluded the PyFPE macros from the limited C API in Python 3.9:
commit 488d02a24142948bfb1fafd19fa48e61fcbbabc5 Author: Victor Stinner \vstinner@python.org\ Date: Wed Nov 20 12:17:09 2019 +0100
bpo-38835: Exclude PyFPE macros from the stable API (GH-17228)
Exclude PyFPE_START_PROTECT() and PyFPE_END_PROTECT() macros of
pyfpe.h from Py_LIMITED_API (stable API).
See also bpo-38835: "pyfpe.h: Exclude PyFPE_START_PROTECT and PyFPE_END_PROTECT from the Py_LIMITED_API".
Using INADA-san recipe https://github.com/methane/notes/tree/master/2020/wchar-cache I found the 62 projects on the PyPI top 4000 which contain "PyFPE" pattern:
asyncpg-0.22.0.tar.gz auditwheel-3.3.1.tar.gz av-8.0.3.tar.gz bcolz-1.2.1.tar.gz BDQuaternions-0.2.15.tar.gz bx-python-0.8.9.tar.gz ConfigSpace-0.4.18.tar.gz Cython-0.29.21.tar.gz cytoolz-0.11.0.tar.gz ddtrace-0.46.0.tar.gz dedupe-hcluster-0.3.8.tar.gz dependency-injector-4.23.5.tar.gz fastavro-1.3.2.tar.gz fastdtw-0.3.4.tar.gz Fiona-1.8.18.tar.gz fuzzysearch-0.7.3.tar.gz fuzzyset-0.0.19.tar.gz gensim-3.8.3.tar.gz grpcio-1.35.0.tar.gz gssapi-1.6.12.tar.gz hdbscan-0.8.27.tar.gz hmmlearn-0.2.5.tar.gz imagecodecs-2021.1.28.tar.gz implicit-0.4.4.tar.gz lightfm-1.16.tar.gz lupa-1.9.tar.gz lxml-4.6.2.tar.gz mujoco-py-2.0.2.13.tar.gz neobolt-1.7.17.tar.gz orderedset-2.0.3.tar.gz peewee-3.14.1.tar.gz Pillow-8.1.0.tar.gz Pillow-SIMD-7.0.0.post3.tar.gz pmdarima-1.8.0.tar.gz pomegranate-0.14.2.tar.gz pycapnp-1.0.0.tar.gz pydevd-2.2.0.tar.gz pygame-2.0.1.tar.gz pyhacrf-datamade-0.2.5.tar.gz pyjq-2.5.1.tar.gz pysam-0.16.0.1.tar.gz pystan-2.19.1.1.tar.gz PyWavelets-1.1.1.tar.gz rasterio-1.2.0.tar.gz ruptures-1.1.3.tar.gz s2sphere-0.2.5.tar.gz scikit-image-0.18.1.tar.gz scipy-1.6.1.tar.gz Shapely-1.7.1.tar.gz simplejson-3.17.2.tar.gz spacy-3.0.3.tar.gz statsmodels-0.12.2.tar.gz tables-3.6.1.tar.gz Theano-1.0.5.tar.gz thinc-8.0.1.tar.gz tinycss-0.4.tar.gz tslearn-0.5.0.5.tar.gz uvloop-0.15.1.tar.gz weighted_levenshtein-0.2.1.tar.gz wordcloud-1.8.1.tar.gz wsaccel-0.6.3.tar.gz wxPython-4.1.1.tar.gz
Example with asyncpg:
$ rg PyFPE
asyncpg/pgproto/pgproto.c
39841: PyFPE_START_PROTECT("add", return NULL)
39843: PyFPE_END_PROTECT(result)
asyncpg/protocol/protocol.c 85260: PyFPE_START_PROTECT("add", return NULL) 85262: PyFPE_END_PROTECT(result) 85563: PyFPE_START_PROTECT("subtract", return NULL) 85565: PyFPE_END_PROTECT(result) 85734: PyFPE_START_PROTECT("add", return NULL) 85736: PyFPE_END_PROTECT(result)
So it doesn't sound like a good idea to immediately remove these two macros.
So it doesn't sound like a good idea to immediately remove these two macros.
MAYBE we can update the annotation in pyfpe.h to address that we will remove those macros in python 4.0.
The fpe have been removed, so those macros will be removed finally.
If I remember correctly, cython-generated C code calls those macros at appropriate places, so there are probably a bunch of projects that are using them and the developers have no idea.
On Thu, Feb 18, 2021, 04:16 STINNER Victor \report@bugs.python.org\ wrote:
STINNER Victor \vstinner@python.org\ added the comment:
Using INADA-san recipe https://github.com/methane/notes/tree/master/2020/wchar-cache I found the 62 projects on the PyPI top 4000 which contain "PyFPE" pattern:
asyncpg-0.22.0.tar.gz auditwheel-3.3.1.tar.gz av-8.0.3.tar.gz bcolz-1.2.1.tar.gz BDQuaternions-0.2.15.tar.gz bx-python-0.8.9.tar.gz ConfigSpace-0.4.18.tar.gz Cython-0.29.21.tar.gz cytoolz-0.11.0.tar.gz ddtrace-0.46.0.tar.gz dedupe-hcluster-0.3.8.tar.gz dependency-injector-4.23.5.tar.gz fastavro-1.3.2.tar.gz fastdtw-0.3.4.tar.gz Fiona-1.8.18.tar.gz fuzzysearch-0.7.3.tar.gz fuzzyset-0.0.19.tar.gz gensim-3.8.3.tar.gz grpcio-1.35.0.tar.gz gssapi-1.6.12.tar.gz hdbscan-0.8.27.tar.gz hmmlearn-0.2.5.tar.gz imagecodecs-2021.1.28.tar.gz implicit-0.4.4.tar.gz lightfm-1.16.tar.gz lupa-1.9.tar.gz lxml-4.6.2.tar.gz mujoco-py-2.0.2.13.tar.gz neobolt-1.7.17.tar.gz orderedset-2.0.3.tar.gz peewee-3.14.1.tar.gz Pillow-8.1.0.tar.gz Pillow-SIMD-7.0.0.post3.tar.gz pmdarima-1.8.0.tar.gz pomegranate-0.14.2.tar.gz pycapnp-1.0.0.tar.gz pydevd-2.2.0.tar.gz pygame-2.0.1.tar.gz pyhacrf-datamade-0.2.5.tar.gz pyjq-2.5.1.tar.gz pysam-0.16.0.1.tar.gz pystan-2.19.1.1.tar.gz PyWavelets-1.1.1.tar.gz rasterio-1.2.0.tar.gz ruptures-1.1.3.tar.gz s2sphere-0.2.5.tar.gz scikit-image-0.18.1.tar.gz scipy-1.6.1.tar.gz Shapely-1.7.1.tar.gz simplejson-3.17.2.tar.gz spacy-3.0.3.tar.gz statsmodels-0.12.2.tar.gz tables-3.6.1.tar.gz Theano-1.0.5.tar.gz thinc-8.0.1.tar.gz tinycss-0.4.tar.gz tslearn-0.5.0.5.tar.gz uvloop-0.15.1.tar.gz weighted_levenshtein-0.2.1.tar.gz wordcloud-1.8.1.tar.gz wsaccel-0.6.3.tar.gz wxPython-4.1.1.tar.gz
Example with asyncpg:
$ rg PyFPE asyncpg/pgproto/pgproto.c 39841: PyFPE_START_PROTECT("add", return NULL) 39843: PyFPE_END_PROTECT(result)
asyncpg/protocol/protocol.c 85260: PyFPE_START_PROTECT("add", return NULL) 85262: PyFPE_END_PROTECT(result) 85563: PyFPE_START_PROTECT("subtract", return NULL) 85565: PyFPE_END_PROTECT(result) 85734: PyFPE_START_PROTECT("add", return NULL) 85736: PyFPE_END_PROTECT(result)
So it doesn't sound like a good idea to immediately remove these two macros.
----------
Python tracker \report@bugs.python.org\ \https://bugs.python.org/issue43250\
I don't have the bandwidth to update Cython and mark the macro as deprecated. So I just close the issue.
I don't have the bandwidth to update Cython and mark the macro as deprecated. So I just close the issue.
I'm about to kill them off in Cython - https://github.com/cython/cython/pull/5841.
Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.
Show more details
GitHub fields: ```python assignee = None closed_at = None created_at =
labels = ['expert-C-API', '3.10']
title = '[C API] Deprecate or remove PyFPE_START_PROTECT() and PyFPE_END_PROTECT()'
updated_at =
user = 'https://github.com/vstinner'
```
bugs.python.org fields:
```python
activity =
actor = 'corona10'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['C API']
creation =
creator = 'vstinner'
dependencies = []
files = []
hgrepos = []
issue_num = 43250
keywords = []
message_count = 8.0
messages = ['387213', '387215', '387218', '387219', '387220', '387225', '387461', '387467']
nosy_count = 4.0
nosy_names = ['vstinner', 'njs', 'corona10', 'shihai1991']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = None
url = 'https://bugs.python.org/issue43250'
versions = ['Python 3.10']
```