gh-125115 : Refactor the pdb parsing issue so positional arguments can pass through

[gaogaotiantian]

We are at a point that we need to do customize parsing for pdb in order to make the positional argument passing intuitive.

For example

# -m 1 should be passed
python -m pdb -m calendar -m 1
# -p 1 should be passed
python -m pdb -m foo -p 3
# -p 1 should be passed
python -m pdb foo.py -p 1

The order of the arguments now matter (especially when we use a very common option -p).

Now pdb try to

1. Extract all -c commands (for backward compatible reason it still get commands after the script/module).
2. Locate the target based on order (the first one except for -c).
3. Consider the rest of the arguments to be positional arguments to the script/module itself.

• Issue: pdb: can't debug a script with arguments #125115

[gaogaotiantian]

Hey @iritkatriel , I think it's time to do some customized parsing for pdb, because we want to resolve the ambiguity for the users. LMK if you have any questions. I added backport to 3.14 because this is indeed kind of a bug. Especially considering 3.14 introduced -p which could conflict a lot.

[iritkatriel]

Is -p always the first arg?

[gaogaotiantian]

So we already skimmed all the -c commands, which are the only "valid options" to all three kinds of invocation. Then we need to determine which path (script, module or pid) we should take. There's no case where -p is not the first argument (after skimming -c) but we should still consider a pid attach.

[iritkatriel]

Probably worth getting a review from @savannahostrowski as the argparse maintainer.

[savannahostrowski]

I think custom parsing logic is necessary here, given that parse_known_args is too greedy and REMAINDER is plagued with issues (see #58382 as an example). There's no combination of argparse features that will make this work, AFAIK.

On a more meta note, it does make me wonder whether it'd be worth exposing a proper passthrough parameter in argparse itself to simplify things, especially since this scenario is actually not that uncommon.

[gaogaotiantian]

Hi @savannahostrowski and @iritkatriel , any more comments on this PR?

[miss-islington-app]

Thanks @gaogaotiantian for the PR 🌮🎉.. I'm working now to backport this PR to: 3.14.
🐍🍒⛏🤖

[miss-islington-app]

Sorry, @gaogaotiantian, I could not cleanly backport this to 3.14 due to a conflict.
Please backport using cherry_picker on command line.

cherry_picker 5348c200f5b26d6dd21d900b2b4cb684150d4b01 3.14

[bedevere-app]

GH-141635 is a backport of this pull request to the 3.14 branch.