Right, that's the bug we found. The statements are in the reverse order. It's not quite that simple as the code below, but it ends up happening in that order.
It's one of the most fundamental things a senior dev asks themselves on every call that that differentiates them from a junior dev - "What is the lifetime semantics of this thing I'm handing over / getting back?". You don't even need the documentation for this one after looking at how OpenSSL gets the socket handle in the first place - the bug is obvious, I just wasn't aware this codepath even existed. (Big product, big team - you know how it goes). - Deon -----Original Message----- From: sqlite-users <sqlite-users-boun...@mailinglists.sqlite.org> On Behalf Of Olivier Mascia Sent: Tuesday, April 17, 2018 2:28 PM To: SQLite mailing list <sqlite-users@mailinglists.sqlite.org> Subject: Re: [sqlite] Header corruption > Le 17 avr. 2018 à 22:07, Deon Brewis <de...@outlook.com> a écrit : > > closesocket(_socket); // posix socket > SSL_shutdown(_ssl); // openssl (_ssl was initialized using the _socket > above) These two statements are inherently wrong, in this order. First you SSL_shutdown(), then you closesocket(). Not the reverse. And in Windows code it even is good citizenship to call shutdown() properly in between. AFAIK it should also be on Linux, though I'm told it is not commonly seen. But please check the respective OS'es SDK and OpenSSL documentation for the details. Called in the right order there is no file descriptor re-use syndrome to fear, because the closesocket() rightfully comes as the very last step. -- Best regards, Meilleures salutations, Met vriendelijke groeten, Olivier Mascia (from mobile device) _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
