You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm currently investigating the issue mentioned here and I've narrowed the situation down somewhat:
SimpleAsyncHTTPRequest is making a POST request with an empty body to a server that is responding very quickly. In this case, it is the request to github.com to complete GitHub OAuth.
When read_until_regex is called in
read buffer is not empty, but
stream is closed
So a closed check is apparently occurring before the read is attempted from the existing buffer. Interestingly, #2670 introduces exactly such a premature closed check (a _try_inline_read() would succeed if attempted), but the issue still occurs in 6.0.2, prior to #2670, so there must be an additional check that's causing StreamClosedError to be raised even when the
I'm trying to provoke this with a self-contained test case that doesn't require github credentials, but haven't managed it yet. Will report back when I do.
Using curl_httpclient does not appear to exhibit this behavior.
I think this is related to #2651 in that it's another issue in 6.0 where there are races between the closed state and pending reads/writes.
The text was updated successfully, but these errors were encountered:
I'm currently investigating the issue mentioned here and I've narrowed the situation down somewhat:
SimpleAsyncHTTPRequest is making a POST request with an empty body to a server that is responding very quickly. In this case, it is the request to github.com to complete GitHub OAuth.
When
read_until_regex
is called inSo a closed check is apparently occurring before the read is attempted from the existing buffer. Interestingly, #2670 introduces exactly such a premature closed check (a
_try_inline_read()
would succeed if attempted), but the issue still occurs in 6.0.2, prior to #2670, so there must be an additional check that's causing StreamClosedError to be raised even when theI'm trying to provoke this with a self-contained test case that doesn't require github credentials, but haven't managed it yet. Will report back when I do.
Using curl_httpclient does not appear to exhibit this behavior.
I think this is related to #2651 in that it's another issue in 6.0 where there are races between the closed state and pending reads/writes.
The text was updated successfully, but these errors were encountered: