I modified Py_Main() to ignore the PYTHONCOERCECLOCALE environment variable if -E or -I command line option is used. But Nick asks to always read PYTHONCOERCECLOCALE.

We should either update the PEP or change the code.

I am not sure why PYTHONCOERCECLOCALE should be handled differently than other PYTHON* variable like PYTHONWARNINGS or PYTHONUTF8. Is it because it impacts the encodings? Is it because there was a chicken-and-egg issue before I reworked Py_Main() code? (PYTHONCOERCECLOCALE env var was read before reading command line arguments.)

--

Copy of Nick Coghlan's msg325009:

(The one exception to "nothing gets decoded incorrectly" is that PYTHONCOERCECLOCALE itself is always interpreted as an ASCII field: the values that it checks for are actually ASCII byte sequences, not Unicode code points.

The documentation could definitely be much clearer on that point though, as even in the PEP it's only implied by the final paragraph in https://www.python.org/dev/peps/pep-0538/#legacy-c-locale-coercion-in-the-standalone-python-interpreter-binary which is mostly talking about the fact that this particular environment variable is still checked, even if you pass the -I or -E command line options.

Respecting -E and -I isn't a problem per se - the problem is moving the _Py_CoerceLegacyLocale call to a point that's incredibly late in the startup process *just* to get it to respect those flags.

I don't actually mind if you reinstate the extra pass through the command line arguments just to check for -E and -I early enough for it to affect a properly located call to _Py_CoerceLegacyLocale, I just don't think it's necessary to do so.

For the PYTHONCOERCECLOCALE=warn case, it turns out that my preferred approach to implementing bpo-34589 also naturally ends up respecting -I and -E for that (i.e. supplying -I or -E will suppressed the warning).

However, my upcoming PR for that also reinstates and expands on my original comment that explained why it was valuable to support "PYTHONCOERCECLOCALE=0 python3 -E ..." and "PYTHONCOERCECLOCALE=0 python3 -I ...": so you can readily reproduce the way that locale coercion behaves on a platform *without* a suitable target locale (e.g. CentOS 7), even if your current platform actually does have such a locale available (e.g. Fedora).

Respecting -E and -I isn't a problem per se - the problem is moving the _Py_CoerceLegacyLocale call to a point that's incredibly late in the startup process *just* to get it to respect those flags.

Would you mind to elaborate how it is an issue? The LC_CTYPE is coerced *before* the configuration is read a second time. If there is an issue, would you mind to show an example where something is decoded from the wrong encoding?

The discussion moved back to bpo-34589.