gh-120606: Allow EOF to exit pdb commands definition

[gaogaotiantian]

• Issue: Allow EOF to exit pdb commands definition #120606

[iritkatriel]

(1) why do we need two instructions to do the same thing?
(2) why add a feature and not document it?
(3) where is the test?

[gaogaotiantian]

(1) Because EOF is a special command triggered when user "inputs" EOF (Ctrl+D on Linux). It's not something we added, it's something we need to deal with. Users will do Ctrl+D as a habit from all the other similar usages and expect a reasonable result. Currently the behavior is pretty bad.

(Pdb) commands 1
(com) (com) (com) (com) (com) (com) 

(2) Like I said, this is a reaction to a very natural user input. I think that's a very expected behavior. However, we can add something in commands command saying that inputting an EOF character will also exit the commands command.

(3) The test involving EOF is not easy to test, especially if you still want to input something after. Currently pdb does not really test it. One way around this is to use "EOF" as an explicit command to simulate the EOF user inputs. That would make the test super trivial to write. We are relying on cmd.Cmd to convert Ctrl+D (Ctrl+Z on Windows I believe) to the correct EOF command. Does that sound reasonable to you?

[iritkatriel]

I see. That context can be encoded in a test. Add the trivial test for 'EOF' and put a comment there that this is what cmd produces for Ctrl+D.

[gaogaotiantian]

Okay I added a basic test in an existing test case. Notice that <BLANKLINE> appears in the output because EOF\n creates a new line but Ctrl-D will not have that new line. So EOF\n results in an extra new line compared to Ctrl-D. We don't want the user to stay on the same line after Ctrl-D so we need to output a new line in that command.

[gaogaotiantian]

Thanks for the review!