Server replay with POST

I have a bunch of captured flows that include some POST requests to the same URL with different bodies, and corresponding different responses.

I want to replay these using -S in an offline scenario. However, if I do not specify the --replay-ignore-content header the requests are not matched (so connection is killed because I use -k to enforce that it’s offline only because I don’t want to hit the third-party server while testing this) and if I do specify --replay-ignore-content then all of the POST requests return the same response (the first one in the file) because the content isn’t included in the match.

Is there a way to see what isn’t matching correctly in the first scenario so I can have different POSTs to the same URL? So far I have verified the method, scheme, host, port, path, and query string, and the content appears to be correct as well.

I am not specifying --norefresh, --server-replay-use-header, etc. I am specifying --no-pop so that I don’t have to restart the proxy to return the same content multiple times. (I want it to vary by the request content)

I have stopped using --no-pop for now so that I can progress, but this results in the server replaying the messages in the order they were originally captured, regardless of the message body. This is less than ideal from a testing standpoint since any POST to a particular route pops the next response message, resulting in there being no relationship between what is posted and what is returned.

Hi @galactic_cowboy,

Sorry for the late reply - if you want to find out what’s different, I would recommend adding a print(key) here: https://github.com/mitmproxy/mitmproxy/blob/2.0.x/mitmproxy/addons/serverplayback.py#L75

Thanks, that’s pretty much exactly what I ended up doing. (Logged to file, but same concept.)

Turned out that it was my own fault. I had translated the dump file to json, thought I had made some modifications, then back to tnetstrings. However, for these forms it was using the original data instead of my modifications because I forgot to change a flag in my software that translated it back and forth. Once I realized that, I was able to test with the original data to prove it worked, then actually make my modifications and test that it still worked.

So everything seems to work as expected now.