We're no longer using _Py_IDENTIFIER() (or _Py_static_string()) in any core CPython code. It is still used in a number of non-builtin stdlib modules.
The replacement is: PyUnicodeObject (not pointer) fields under _PyRuntimeState, statically initialized as part of _PyRuntime. A new _Py_GET_GLOBAL_IDENTIFIER() macro facilitates lookup of the fields (along with _Py_GET_GLOBAL_STRING() for non-identifier strings).
https://bugs.python.org/issue46541#msg411799 explains the rationale for this change.
The core of the change is in:
* (new) Include/internal/pycore_global_strings.h - the declarations for the global strings, along with the macros
* Include/internal/pycore_runtime_init.h - added the static initializers for the global strings
* Include/internal/pycore_global_objects.h - where the struct in pycore_global_strings.h is hooked into _PyRuntimeState
* Tools/scripts/generate_global_objects.py - added generation of the global string declarations and static initializers
I've also added a --check flag to generate_global_objects.py (along with make check-global-objects) to check for unused global strings. That check is added to the PR CI config.
The remainder of this change updates the core code to use _Py_GET_GLOBAL_IDENTIFIER() instead of _Py_IDENTIFIER() and the related _Py*Id functions (likewise for _Py_GET_GLOBAL_STRING() instead of _Py_static_string()). This includes adding a few functions where there wasn't already an alternative to _Py*Id(), replacing the _Py_Identifier * parameter with PyObject *.
The following are not changed (yet):
* stop using _Py_IDENTIFIER() in the stdlib modules
* (maybe) get rid of _Py_IDENTIFIER(), etc. entirely -- this may not be doable as at least one package on PyPI using this (private) API
* (maybe) intern the strings during runtime init
https://bugs.python.org/issue46541
Prepare for module state:
- Add "get state by defining class" and "get state by module def" stubs
- Add AC defining class when needed
- Add state pointer to connection context
- Pass state as argument to utility functions
Automerge-Triggered-By: GH:encukou
With this, all sqlite3 static globals have been moved to the global state.
There are a couple of global static strings left, but there should be no need for adding them to the state.
https://bugs.python.org/issue42064
* Move connection type to global state
* Move cursor type to global state
* Move prepare protocol type to global state
* Move row type to global state
* Move statement type to global state
* ADD_TYPE takes a pointer
* pysqlite_get_state is now static inline
The bulk of this patch was generated automatically with:
for name in \
PyObject_Vectorcall \
Py_TPFLAGS_HAVE_VECTORCALL \
PyObject_VectorcallMethod \
PyVectorcall_Function \
PyObject_CallOneArg \
PyObject_CallMethodNoArgs \
PyObject_CallMethodOneArg \
;
do
echo $name
git grep -lwz _$name | xargs -0 sed -i "s/\b_$name\b/$name/g"
done
old=_PyObject_FastCallDict
new=PyObject_VectorcallDict
git grep -lwz $old | xargs -0 sed -i "s/\b$old\b/$new/g"
and then cleaned up:
- Revert changes to in docs & news
- Revert changes to backcompat defines in headers
- Nudge misaligned comments
svn+ssh://pythondev@svn.python.org/python/trunk
........
r66394 | benjamin.peterson | 2008-09-11 17:04:02 -0500 (Thu, 11 Sep 2008) | 1 line
fix typo
........
r66404 | gerhard.haering | 2008-09-12 08:54:06 -0500 (Fri, 12 Sep 2008) | 2 lines
sqlite3 module: Mark iterdump() method as "Non-standard" like all the other methods not found in DB-API.
........
r66412 | gerhard.haering | 2008-09-12 13:58:57 -0500 (Fri, 12 Sep 2008) | 2 lines
Fixes issue #3103. In the sqlite3 module, made one more function static. All renaming public symbos now have the pysqlite prefix to avoid name clashes. This at least once created problems where the same symbol name appeared somewhere in Apache and the sqlite3 module was used from mod_python.
........
r66414 | gerhard.haering | 2008-09-12 17:33:22 -0500 (Fri, 12 Sep 2008) | 2 lines
Issue #3846: Release GIL during calls to sqlite3_prepare. This improves concurrent access to the same database file from multiple threads/processes.
........
r66424 | andrew.kuchling | 2008-09-12 20:22:08 -0500 (Fri, 12 Sep 2008) | 1 line
#687648 from Robert Schuppenies: use classic division. (RM Barry gave permission to update the demos.)
........
r66425 | andrew.kuchling | 2008-09-12 20:27:33 -0500 (Fri, 12 Sep 2008) | 1 line
#687648 from Robert Schuppenies: use classic division. From me: don't use string exception; flush stdout after printing
........
r66426 | andrew.kuchling | 2008-09-12 20:34:41 -0500 (Fri, 12 Sep 2008) | 1 line
#687648 from Robert Schuppenies: use classic division. From me: don't use string exception; add __main__ section
........
r66427 | andrew.kuchling | 2008-09-12 20:42:55 -0500 (Fri, 12 Sep 2008) | 1 line
#687648 from Robert Schuppenies: use classic division. From me: remove two stray semicolons
........
r66428 | andrew.kuchling | 2008-09-12 20:43:28 -0500 (Fri, 12 Sep 2008) | 1 line
#687648 from Robert Schuppenies: use classic division.
........
r66429 | andrew.kuchling | 2008-09-12 20:47:02 -0500 (Fri, 12 Sep 2008) | 1 line
Remove semicolon
........
r66430 | andrew.kuchling | 2008-09-12 20:48:36 -0500 (Fri, 12 Sep 2008) | 1 line
Subclass exception
........
r66431 | andrew.kuchling | 2008-09-12 20:56:56 -0500 (Fri, 12 Sep 2008) | 1 line
Fix SyntaxError
........
r66432 | andrew.kuchling | 2008-09-12 20:57:25 -0500 (Fri, 12 Sep 2008) | 1 line
Update uses of string exceptions
........
r66433 | andrew.kuchling | 2008-09-12 21:08:30 -0500 (Fri, 12 Sep 2008) | 1 line
Use title case
........
r66434 | andrew.kuchling | 2008-09-12 21:09:15 -0500 (Fri, 12 Sep 2008) | 1 line
Remove extra 'the'; the following title includes it
........
r66435 | andrew.kuchling | 2008-09-12 21:11:51 -0500 (Fri, 12 Sep 2008) | 1 line
#3288: Document as_integer_ratio
........
r66436 | andrew.kuchling | 2008-09-12 21:14:15 -0500 (Fri, 12 Sep 2008) | 1 line
Use title case
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r53304 | vinay.sajip | 2007-01-09 15:50:28 +0100 (Tue, 09 Jan 2007) | 1 line
Bug #1627575: Added _open() method to FileHandler which can be used to reopen files. The FileHandler instance now saves the encoding (which can be None) in an attribute called "encoding".
........
r53305 | vinay.sajip | 2007-01-09 15:51:36 +0100 (Tue, 09 Jan 2007) | 1 line
Added entry about addition of _open() method to logging.FileHandler.
........
r53306 | vinay.sajip | 2007-01-09 15:54:56 +0100 (Tue, 09 Jan 2007) | 1 line
Added a docstring
........
r53316 | thomas.heller | 2007-01-09 20:19:33 +0100 (Tue, 09 Jan 2007) | 4 lines
Verify the sizes of the basic ctypes data types against the struct
module.
Will backport to release25-maint.
........
r53340 | gustavo.niemeyer | 2007-01-10 17:13:40 +0100 (Wed, 10 Jan 2007) | 3 lines
Mention in the int() docstring that a base zero has meaning, as
stated in http://docs.python.org/lib/built-in-funcs.html as well.
........
r53341 | gustavo.niemeyer | 2007-01-10 17:15:48 +0100 (Wed, 10 Jan 2007) | 2 lines
Minor change in int() docstring for proper spacing.
........
r53358 | thomas.heller | 2007-01-10 21:12:13 +0100 (Wed, 10 Jan 2007) | 1 line
Change the ctypes version number to "1.1.0".
........
r53361 | thomas.heller | 2007-01-10 21:51:19 +0100 (Wed, 10 Jan 2007) | 1 line
Must change the version number in the _ctypes extension as well.
........
r53362 | guido.van.rossum | 2007-01-11 00:12:56 +0100 (Thu, 11 Jan 2007) | 3 lines
Fix the signature of log_error(). (A subclass that did the right thing
was getting complaints from pychecker.)
........
r53370 | matthias.klose | 2007-01-11 11:26:31 +0100 (Thu, 11 Jan 2007) | 2 lines
- Make the documentation match the code and the docstring
........
r53375 | matthias.klose | 2007-01-11 12:44:04 +0100 (Thu, 11 Jan 2007) | 2 lines
- idle: Honor the "Cancel" action in the save dialog (Debian bug #299092).
........
r53381 | raymond.hettinger | 2007-01-11 19:22:55 +0100 (Thu, 11 Jan 2007) | 1 line
SF #1486663 -- Allow keyword args in subclasses of set() and frozenset().
........
r53388 | thomas.heller | 2007-01-11 22:18:56 +0100 (Thu, 11 Jan 2007) | 4 lines
Fixes for 64-bit Windows: In ctypes.wintypes, correct the definitions
of HANDLE, WPARAM, LPARAM data types. Make parameterless foreign
function calls work.
........
r53390 | thomas.heller | 2007-01-11 22:23:12 +0100 (Thu, 11 Jan 2007) | 2 lines
Correct the comments: the code is right.
........
r53393 | brett.cannon | 2007-01-12 08:27:52 +0100 (Fri, 12 Jan 2007) | 3 lines
Fix error where the end of a funcdesc environment was accidentally moved too
far down.
........
r53397 | anthony.baxter | 2007-01-12 10:35:56 +0100 (Fri, 12 Jan 2007) | 3 lines
add parsetok.h as a dependency - previously, changing this file doesn't
cause the right files to be rebuilt.
........
r53401 | thomas.heller | 2007-01-12 21:08:19 +0100 (Fri, 12 Jan 2007) | 3 lines
Avoid warnings in the test suite because ctypes.wintypes cannot be
imported on non-windows systems.
........
r53402 | thomas.heller | 2007-01-12 21:17:34 +0100 (Fri, 12 Jan 2007) | 6 lines
patch #1610795: BSD version of ctypes.util.find_library, by Martin
Kammerhofer.
release25-maint backport candidate, but the release manager has to
decide.
........
r53403 | thomas.heller | 2007-01-12 21:21:53 +0100 (Fri, 12 Jan 2007) | 3 lines
patch #1610795: BSD version of ctypes.util.find_library, by Martin
Kammerhofer.
........
r53406 | brett.cannon | 2007-01-13 01:29:49 +0100 (Sat, 13 Jan 2007) | 2 lines
Deprecate the sets module.
........
r53407 | georg.brandl | 2007-01-13 13:31:51 +0100 (Sat, 13 Jan 2007) | 3 lines
Fix typo.
........
r53409 | marc-andre.lemburg | 2007-01-13 22:00:08 +0100 (Sat, 13 Jan 2007) | 16 lines
Bump version number and change copyright year.
Add new API linux_distribution() which supports reading the full distribution
name and also knows how to parse LSB-style release files.
Redirect the old dist() API to the new API (using the short distribution name
taken from the release file filename).
Add branch and revision to _sys_version().
Add work-around for Cygwin to libc_ver().
Add support for IronPython (thanks for Anthony Baxter) and make
Jython support more robust.
........
r53410 | neal.norwitz | 2007-01-13 22:22:37 +0100 (Sat, 13 Jan 2007) | 1 line
Fix grammar in docstrings
........
r53411 | marc-andre.lemburg | 2007-01-13 23:32:21 +0100 (Sat, 13 Jan 2007) | 9 lines
Add parameter sys_version to _sys_version().
Change the cache for _sys_version() to take the parameter into account.
Add support for parsing the IronPython 1.0.1 sys.version value - even
though it still returns '1.0.0'; the version string no longer includes
the patch level.
........
r53412 | peter.astrand | 2007-01-13 23:35:35 +0100 (Sat, 13 Jan 2007) | 1 line
Fix for bug #1634343: allow specifying empty arguments on Windows
........
r53414 | marc-andre.lemburg | 2007-01-13 23:59:36 +0100 (Sat, 13 Jan 2007) | 14 lines
Add Python implementation to the machine details.
Pretty-print the Python version used for running PyBench.
Let the user know when calibration has finished.
[ 1563844 ] pybench support for IronPython:
Simplify Unicode version detection.
Make garbage collection and check interval settings optional if
the Python implementation doesn't support thess (e.g. IronPython).
........
r53415 | marc-andre.lemburg | 2007-01-14 00:13:54 +0100 (Sun, 14 Jan 2007) | 5 lines
Use defaults if sys.executable isn't set (e.g. on Jython).
This change allows running PyBench under Jython.
........
r53416 | marc-andre.lemburg | 2007-01-14 00:15:33 +0100 (Sun, 14 Jan 2007) | 3 lines
Jython doesn't have sys.setcheckinterval() - ignore it in that case.
........
r53420 | gerhard.haering | 2007-01-14 02:43:50 +0100 (Sun, 14 Jan 2007) | 29 lines
Merged changes from standalone version 2.3.3. This should probably all be
merged into the 2.5 maintenance branch:
- self->statement was not checked while fetching data, which could
lead to crashes if you used the pysqlite API in unusual ways.
Closing the cursor and continuing to fetch data was enough.
- Converters are stored in a converters dictionary. The converter name
is uppercased first. The old upper-casing algorithm was wrong and
was replaced by a simple call to the Python string's upper() method
instead.
-Applied patch by Glyph Lefkowitz that fixes the problem with
subsequent SQLITE_SCHEMA errors.
- Improvement to the row type: rows can now be iterated over and have a keys()
method. This improves compatibility with both tuple and dict a lot.
- A bugfix for the subsecond resolution in timestamps.
- Corrected the way the flags PARSE_DECLTYPES and PARSE_COLNAMES are
checked for. Now they work as documented.
- gcc on Linux sucks. It exports all symbols by default in shared
libraries, so if symbols are not unique it can lead to problems with
symbol lookup. pysqlite used to crash under Apache when mod_cache
was enabled because both modules had the symbol cache_init. I fixed
this by applying the prefix pysqlite_ almost everywhere. Sigh.
........
r53423 | guido.van.rossum | 2007-01-14 04:46:33 +0100 (Sun, 14 Jan 2007) | 2 lines
Remove a dependency of this test on $COLUMNS.
........
r53425 | ka-ping.yee | 2007-01-14 05:25:15 +0100 (Sun, 14 Jan 2007) | 3 lines
Handle old-style instances more gracefully (display documentation on
the relevant class instead of documentation on <type 'instance'>).
........
r53440 | vinay.sajip | 2007-01-14 22:49:59 +0100 (Sun, 14 Jan 2007) | 1 line
Added WatchedFileHandler (based on SF patch #1598415)
........
r53441 | vinay.sajip | 2007-01-14 22:50:50 +0100 (Sun, 14 Jan 2007) | 1 line
Added documentation for WatchedFileHandler (based on SF patch #1598415)
........
r53442 | guido.van.rossum | 2007-01-15 01:02:35 +0100 (Mon, 15 Jan 2007) | 2 lines
Doc patch matching r53434 (htonl etc. now always take/return positive ints).
........
number of tests, all because of the codecs/_multibytecodecs issue described
here (it's not a Py3K issue, just something Py3K discovers):
http://mail.python.org/pipermail/python-dev/2006-April/064051.html
Hye-Shik Chang promised to look for a fix, so no need to fix it here. The
tests that are expected to break are:
test_codecencodings_cn
test_codecencodings_hk
test_codecencodings_jp
test_codecencodings_kr
test_codecencodings_tw
test_codecs
test_multibytecodec
This merge fixes an actual test failure (test_weakref) in this branch,
though, so I believe merging is the right thing to do anyway.
Erlend Egeberg Aasland
Eric Snow
Victor Stinner
Petr Viktorin
Jeroen Demeyer
Serhiy Storchaka
Benjamin Peterson
Martin v. Löwis
Thomas Wouters