Repository navigation
[_winapi] correctly sort and remove duplicates in getenvironment() #87868
Description
Activity
The
getenvironment()function in "Modules/_winapi.c" needs to sort variables in the environment block and remove duplicates case insensitively1.The sort order used to matter with
SetEnvironmentVairableW(). As soon as it reached a name in the environment block that compared greater than the target name, it would insert a new variable. Nowadays,SetEnvironmentVairableW()searches the entire environment block before inserting a new value. Regardless, at the very least,getenvironment()is not well-behaved and is not setting the environment in the documented sort order that users, and possibly other programs, expect.Case-insensitive sorting in Windows uses upper case. The variable names in the mapping can be added to a list and sorted with a key function that's based on
LCMapStringEx(), with the flagLCMAP_UPPERCASE. Loop over the sorted list to create the environment block. Remove duplicates by skipping a name that compares equal to the previously stored name according toCompareStringOrdinal().Footnotes
- added3.8 (EOL)end of lifeend of life3.10 (EOL)end of lifeend of life3.9 (EOL)end of lifeend of lifeextension-modulesC modules in the Modules dirC modules in the Modules dirtype-bugAn unexpected behavior, bug, or errorAn unexpected behavior, bug, or error
on Apr 2, 2021 - added3.11only security fixesonly security fixesand removed3.8 (EOL)end of lifeend of life
on Feb 26, 2022 I have a question, how to determine which name should be stored if they are duplicated with case insensitive?
which name should be stored if they are duplicated with case insensitive?
Ideally
os.environwould preserve the original case of the process environment, andos.environ.copy()would return a copy that's also case insensitive. That would prevent most problems with duplicates keys. See msg387676 in bpo-28824, and msg414319 in bpo-15373.In msg390038 I suggested keeping the first key that's encountered. However, dicts preserve insertion order nowadays, so one could assume that the last one is the one that the caller wants to keep.
This issue can be reproduced on msys2 (also cygwin, I think) easily:
msys2 shell $ ABC=1 abc=2 python Python 3.11.0b1 (main, May 7 2022, 22:58:47) [MSC v.1931 64 bit (AMD64)] on win32 Type "help", "copyright", "credits" or "license" for more information. >>> import os >>> print(os.environ["ABC"], os.environ["abc"]) 2 2 >>> import subprocess >>> subprocess.run(["bash", "-c", "echo $ABC $abc"]) 1 2 CompletedProcess(args=['bash', '-c', 'echo $ABC $abc'], returncode=0) >>> >>> os.environ["abc"] = "3" >>> print(os.environ["ABC"], os.environ["abc"]) 3 3 >>> subprocess.run(["bash", "-c", "echo $ABC $abc"]) 3 2 CompletedProcess(args=['bash', '-c', 'echo $ABC $abc'], returncode=0)
msys2 shell $ ABC=1 abc=2 cmd cmd > echo %ABC% %abc% 1 1 cmd > bash -c "echo $ABC $abc" 1 2 cmd > set abc=3 cmd > echo %ABC% %abc% 3 3 cmd > bash -c "echo $ABC $abc" 3 2
msys2 shell $ ABC=1 abc=2 powershell PS > echo $Env:ABC $Env:abc 1 1 PS > bash -c "echo $ABC $abc" 1 2 PS > $Env:abc = 3 PS > echo $Env:ABC $Env:abc 3 3 PS > bash -c 'echo $ABC $abc' 3 2
Python behaves differently from both cmd.exe and powershell.
Case-sensitive environment variables are in
nt.environ, but they don't affect the real environment:msys2 shell $ ABC=1 abc=2 python Python 3.11.0b1 (main, May 7 2022, 22:58:47) [MSC v.1931 64 bit (AMD64)] on win32 Type "help", "copyright", "credits" or "license" for more information. >>> import nt, os >>> print(nt.environ["ABC"], nt.environ["abc"]) 1 2 >>> print(os.environ["ABC"], os.environ["abc"]) 2 2
This is because environment vairiables are not sorted correctly in
os._convertenviron:https://github.com/python/cpython/blob/v3.11.0b1/Lib/os.py#L750
- changed the title
[-][Windows] correctly sort and remove duplicates in _winapi getenvironment()[/-][+][_winapi] correctly sort and remove duplicates in getenvironment()[/+]on Aug 8, 2022 - added a commit that references this issue
on Mar 15, 2023 5 remaining items
- added a commit that references this issue
on Jan 10, 2024 i'll apply the test skip PR into the existing 3.12 and 3.11 backport branches.
Reacted by Steve Dower, Erlend E. Aasland and An Long
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:
bugs.python.org fields:
Linked PRs
test_subprocessif python is configured with --enable-shared #113867