Andrew Svetlov
Andrew Svetlov
The best we can do here is dropping connection without any output (but logging the fact). `BadStatusLine` is very exceptional. In general, it is a kind of parser error. Nothing...
Would you make a pull request?
In aiohttp, we use middlewares for error handling. The fix should follow this way I guess (but still support low-level HTTP server)
The thread starts from asking for handling the error by `on_prepare` signal. Middleware is another way to handle it.
No, `aiohttp.web_exceptions` are not derived from `http.client` ones. There are many reasons for it but the first is: server-side exceptions for 400+ status codes have no common meaning with 400+...
Heh, encrypted SSL channel doesn't look like a regular plain HTTP status line, isn't it?
There is no well-formed request exists, no middleware is affected. The error happens on the HTTP protocol level, even before aiohttp routing starts to work. In other words, middlewares/signals are...
The solution is: aiohttp should not requote HTTP redirect URL. Other quoting logic should be untouched. The issue is easy and wais for a champion.
Hi. There is (not finished yet) #4000 that rewrites WebSockets slightly. I believe it will solve this problem among many others. Stay tuned.