David Röthlisberger wrote: > ISSUE 1. > > For client POST requests with RFC 2388 multipart/form-data content > type, polipo's request to the origin server doesn't seem to be > complete. The origin server complains of "request timeout" after one > minute. > [...] > > ISSUE 2. > > Intermittently, polipo crashes just after receiving the above timeout > response from the server and serving it to the client. I can reproduce > this 80% of the time on OS X, but so far not on Linux. In the test > above, curl does print the 408 response from my testwebserver, so > polipo has served the response before crashing. > [...]
I can confirm both issues on Linux x86_64 (by running Curl's test case 9 against polipo from master branch). Issue 2 has already been reported: http://permalink.gmane.org/gmane.comp.web.polipo.user/2144 Here is a possible fix: http://permalink.gmane.org/gmane.comp.web.polipo.user/2147 But Juliusz doesn't like it and is apparently still thinking about it: http://permalink.gmane.org/gmane.comp.web.polipo.user/2160 In short: Curl (the client) sends an "Expect: 100-continue" header. If the server does not answer, Polipo crashes: After the request has been destroyed httpServerContinueConditionHandler is called, which dereferences connection->request. My analysis of issue 1: Curl does not send the body immediately because it waits for a 100 continue. The test server doesn't answer at all and therefore Curl sends the body anyway. Polipo doesn't pass the body to the Server. It seems to me that Polipo waits for the server because of the "Expect: 100-continue" header. Greetings Urs ------------------------------------------------------------------------------ Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_nov _______________________________________________ Polipo-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/polipo-users
