Fix test_max_incomplete_event_size_countermeasure test

[anton-ryzhov]

This test was not actually testing what it should — max_incomplete_event_size limit is only checked for incomplete events, for request before "\r\n\r\n".

get_all_events() before "\r\n\r\n" conditionally raises an exception depending on max_incomplete_event_size

[anton-ryzhov]

@njsmith does this PR miss something or could it be merged?

[njsmith]

I guess I'm not seeing anything wrong with this, but I also don't understand what the purpose of this is? What was it not testing before that it does test now?

[anton-ryzhov]

Thesetwo blocks demonstrate and test that the same payload is handled fine with max_incomplete_event_size=5000 but raises an error with max_incomplete_event_size=4000.

But in fact the first block doesn't ensure that, it doesn't fail if we set max_incomplete_event_size=1000. But it does fail if we add an assert.

So, the successful case was always successful regardless of the configuration

[njsmith]

Ohh I see yeah that makes sense for the first change, good catch. Still don't think I understand the motivation for the second change?

[anton-ryzhov]

Still don't think I understand the motivation for the second change?

To be honest I don't remember.

• Maybe it was for sake of consistency, to check the same in both (line 428) cases
• Maybe a semantic change, instead of "the very first event raises an error" to "at any step of processing, this error will be triggered"
• Maybe it played some role in this change

Do you want me to revert it?