Repository navigation
PYTHONDONTWRITEBYTECODE and PYTHONOPTIMIZE have no effect #76026
Description
| BPO | 31845 |
|---|---|
| Nosy | @ncoghlan, @ericsnowcurrently, @Perlence |
| PRs |
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:
assignee = 'https://github.com/ncoghlan'
closed_at = <Date 2017-10-25.02:14:21.810>
created_at = <Date 2017-10-23.11:25:30.491>
labels = ['interpreter-core', 'type-bug', '3.7']
title = 'PYTHONDONTWRITEBYTECODE and PYTHONOPTIMIZE have no effect'
updated_at = <Date 2017-11-05.04:58:48.549>
user = 'https://github.com/Perlence'bugs.python.org fields:
activity = <Date 2017-11-05.04:58:48.549>
actor = 'ncoghlan'
assignee = 'ncoghlan'
closed = True
closed_date = <Date 2017-10-25.02:14:21.810>
closer = 'ncoghlan'
components = ['Interpreter Core']
creation = <Date 2017-10-23.11:25:30.491>
creator = 'Perlence'
dependencies = []
files = []
hgrepos = []
issue_num = 31845
keywords = ['patch']
message_count = 8.0
messages = ['304794', '304796', '304798', '304899', '304955', '304956', '305582', '305584']
nosy_count = 3.0
nosy_names = ['ncoghlan', 'eric.snow', 'Perlence']
pr_nums = ['4105', '4286']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue31845'
versions = ['Python 3.7']
Activity
Perlence commentedon Oct 23, 2017
Setting PYTHONDONTWRITEBYTECODE doesn't seem to have an effect in Python 3.7.0a2:
$ python -V Python 3.7.0a2 $ env PYTHONDONTWRITEBYTECODE=1 python -c 'import sys; print(sys.dont_write_bytecode)' FalsePerlence commentedon Oct 23, 2017
Looks like PYTHONOPTIMIZE has no effect either:
$ python -V Python 3.7.0a2 $ env PYTHONOPTIMIZE=1 python -c 'import sys; print(sys.flags.optimize)' 0[-]Envvar PYTHONDONTWRITEBYTECODE is not respected[/-][+]PYTHONDONTWRITEBYTECODE and PYTHONOPTIMIZE have no effect[/+]Perlence commentedon Oct 23, 2017
It seems 1abcf67[1] is the first bad commit.
[1]1abcf67
ncoghlan commentedon Oct 24, 2017
Huh, it appears the tests for these features are relying solely on the command line options, and not checking that the environment variables are also setting the flags properly. We should be able to account for that omission with a single new test in
test_cmd_linethat sets every relevant environment variable and checks for the expectedsys.flagscontents.As far as what's actually going wrong goes, it's this sequence of events in Py_Main:
_Py_InitializeCore is setting the internal flags appropriately based on the runtime environment, but then Py_Main is stomping over those environmental settings with the settings from the command line.
Historically, these operations happened the other way around, so it was solely up to the code reading the environment variables to ensure they played nice with each other. Now the command line processing logic needs to be updated to also ensure that it only ever increases these values and never reduces them.
ncoghlan commentedon Oct 25, 2017
New changeset d7ac061 by Nick Coghlan in branch 'master':
bpo-31845: Fix reading flags from environment (GH-4105)
d7ac061
ncoghlan commentedon Oct 25, 2017
We could still use some more comprehensive test cases for the env var handling and the way that interacts with the command line settings, but the merged PR includes at least a rudimentary check for the four that directly affect sys.flags without any tricky side effects (PYTHONDEBUG, PYTHONVERBOSE, PYTHONOPTIMIZE, PYTHONDONTWRITEBYTECODE).
ncoghlan commentedon Nov 5, 2017
Cross-linking to the work-in-progress RFE that introduced the error: https://bugs.python.org/issue22257
ncoghlan commentedon Nov 5, 2017
New changeset 1b46131 by Nick Coghlan in branch 'master':
bpo-22257: Mention startup refactoring in What's New (GH-4286)
1b46131