`PYTHONINSPECT` is checked inconsistently

[StanFromIreland]

Found when testing PYTHONINSPECT for #143134, when you set PYTHONINSPECT to a "non-empty" string that is 0, it is checked in initconfig.c with _Py_get_env_flag:

cpython/Python/initconfig.c

Line 1849 in fc2f0fe

_Py_get_env_flag(use_env, &config->inspect, "PYTHONINSPECT");

But then directly with _Py_GetEnv in pymain_repl:

cpython/Modules/main.c

Lines 579 to 581 in fc2f0fe

	if (!config->inspect && _Py_GetEnv(config->use_environment, "PYTHONINSPECT")) {
	pymain_set_inspect(config, 1);
	}

So, you end up in this situation:

$ PYTHONINSPECT=0 ./python -c "import sys; print(sys.flags.inspect)"
0
>>> # We still inspect, even though sys.flags.inspect is not set...

Linked PRs

• gh-143135: Fix sys.flags.inspect when PYTHONINSPECT=0 #143136
• [3.14] gh-143135: Fix sys.flags.inspect when PYTHONINSPECT=0 (GH-143136) #143153
• [3.13] gh-143135: Fix sys.flags.inspect when PYTHONINSPECT=0 (GH-143136) #143168

[serhiy-storchaka]

According to documentation:

   If this is set to a non-empty string it is equivalent to specifying the
   :option:`-i` option.

So, sys.flags.inspect is incorrect.

[StanFromIreland]

So you prefer to fix them all: #143134?

[serhiy-storchaka]

No, in other cases the documentation is wrong. But in this case it seems that only the value of sys.flags.inspect is incorrect. The interpreter behavior and the documentation are consistent.