Web: catch OSError crash

[treemo]

2015-06-09 17:16:09 CRITICAL ERROR <handler[*][read] (HTTP._on_read)> (<read[web] (<ssl.SSLSocket [closed] fd=-1, family=AddressFamily.AF_INET, type=2049, proto=0>, b'POST / HTTP/1.1\r\nAccept-Encoding: identity\r\nConnection: close\r\nContent-Length: 4\r\nUser-Agent: Python-urllib/3.4\r\nHost: example.com\r\nContent-Type: application/x-www-form-urlencoded\r\n\r\nlog=test' )>) (<class 'OSError'>): OSError(9, 'Bad file descriptor') File "circuits/circuits/core/manager.py", line 668, in _dispatcher
value = event_handler(_eargs, *_ekwargs)
File "circuits/circuits/web/http.py", line 270, in _on_read
headers=parser.get_headers(), server=self._server
File "circuits/circuits/web/wrappers.py", line 186, in init
name = sock.getsockname()

[landscape-bot]

Code quality remained the same when pulling 7ce3aab on treemo:master into 16d52a1 on circuits:master.

[prologic]

Erroneous patch? :)

[treemo]

this patch does not resolve the source of the problem (multiple OSError appears in circuits, when web ssl are used) ^^

[prologic]

Ahh got a traceback and something to report as an issue?

[treemo]

OsError appear random (one part of the web requests working properly):

2015-06-09 17:16:09 CRITICAL ERROR (, b'POST / HTTP/1.1\r\nAccept-Encoding: identity\r\nConnection: close\r\nContent-Length: 4\r\nUser-Agent: Python-urllib/3.4\r\nHost: example.com\r\nContent-Type: application/x-www-form-urlencoded\r\n\r\nlog=test' )>) (): OSError(9, 'Bad file descriptor') File "/usr/local/lib/python3.4/dist-packages/circuits-dev-py3.4.egg/circuits/circuits/core/manager.py", line 668, in dispatcher
value = event_handler(eargs, *ekwargs)
  File "/usr/local/lib/python3.4/dist-packages/circuits-dev-py3.4.egg/circuits/circuits/web/http.py", line 270, in on_read
    headers=parser.get_headers(), server=self._server
  File "/usr/local/lib/python3.4/dist-packages/circuits-dev-py3.4.egg/circuits/circuits/web/wrappers.py", line 186, in __init
    name = sock.getsockname()

2015-06-09 19:09:14 CRITICAL ERROR <handler[*][_read][1.00] (TCPServer._on_read)> (<_read[web] (<socket.socket fd=4, family=AddressFamily.AF_INET, type=2049, proto=0, laddr=('0.0.0.0', 443)> )>) (<class 'OSError'>): OSError(9, 'Bad file descriptor')  File "/usr/local/lib/python3.4/dist-packages/circuits-dev-py3.4.egg/circuits/core/manager.py", line 668, in _dispatcher
    value = event_handler(*eargs, **ekwargs)
  File "/usr/local/lib/python3.4/dist-packages/circuits-dev-py3.4.egg/circuits/net/sockets.py", line 584, in _on_read
    self._accept()
  File "/usr/local/lib/python3.4/dist-packages/circuits-dev-py3.4.egg/circuits/net/sockets.py", line 541, in _accept
    newsock.shutdown(2)

[prologic]

Hmm is this a case of the peer has shutdown it's end mid-way through the request/response?

[treemo]

that's what I suspect

[treemo]

when reverse proxy used: it work

INTERNET <=> circuit (web with ssl): OsError appear random
INTERNET <=> nginx (with ssl) <=> circuit (web without ssl): it work

[prologic]

Ahh interesting :)

[spaceone]

I don't quite understand when this happens. Can you explain it to me?
Well, socket.getsockname() may raise if the socket is not anymore connected.
Does it have to do with the reverse proxy you are using?

[treemo]

@spaceone this error appear randomly (source unknown)
the socket are apparently closed (I don't know why) only if SSL is used on Web server
the reverse proxy fix this issue (apparently the problem is in the Web SSL server)