This hunk in drupal_http_request() always executes at least twice, because $info['eof'] will be NULL whatever happens during the first request (unless it's actually empty).
It looks like we could use $info['unread_bytes'] instead, and save the second request just to find out we got to the end of the file.
Testing with openid, I didn't see any bugs with this, but let's see what the testbot says.
Found this via profiling both a recieving openid_provider, and a sending openid module on two separate servers.
| Comment | File | Size | Author |
|---|---|---|---|
| drupal_http_request.patch | 627 bytes | catch |
Comments
Comment #2
tr commenteddrupal_http_request.patch queued for re-testing.
Comment #4
mikeytown2 commentedThe documentation says that you shouldn't use 'unread_bytes' in stream_get_meta_data(). I was looking at a way to get around this and there is no good way to do it. Coming from #1320222: Bring in other drupal patches if your wondering.
Comment #5
catchI think this is probably not worth worrying about. The performance issue I was debugging turned out to be the openid math library used at the other side of the connection.
Comment #6
mikeytown2 commentedThinking about this, if chunked transfer encoding is used then you can detect that the stream is done because the last chunk is a zero-length chunk http://en.wikipedia.org/wiki/Chunked_transfer_encoding#Format; Once you see a chunk size of 0 close the connection. One could also close the connection after the Content-Length has been hit.
#106506: drupal_http_request() does not handle 'chunked' responses - Make it support HTTP 1.1