[openssl.org #1591] get_session_cb callback invoked with no previous session in 0.9.8f

2007-10-19 Thread Lutz Jaenicke via RT
Fixed in 0.9.8g __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager [EMAIL PROTECTED]

[openssl.org #1591] get_session_cb callback invoked with no previous session in 0.9.8f

2007-10-17 Thread [EMAIL PROTECTED] via RT
Starting with OpenSSL 0.9.8f, ssl3_get_client_hello() no longer tests whether the client proposed a previous session_id before trying to process it. In previous releases, a new session was always created if no previous session was proposed (i.e. if j==0 at ssl\s3_srvr.c:746)

[openssl.org #1591] get_session_cb callback invoked with no previous session in 0.9.8f

2007-10-17 Thread Lutz Jaenicke via RT
[EMAIL PROTECTED] - Wed Oct 17 18:11:27 2007]: Starting with OpenSSL 0.9.8f, ssl3_get_client_hello() no longer tests whether the client proposed a previous session_id before trying to process it. In previous releases, a new session was always created if no previous session was proposed

Re: [openssl.org #1591] get_session_cb callback invoked with no previous session in 0.9.8f

2007-10-17 Thread [EMAIL PROTECTED] via RT
Hi Lutz, Apologies, I should have included a stack trace with the bug report. FYI - attached is a Windows/Apache 2.2.6/OpenSSL 0.9.8f VC8 stack trace. The problem is not Windows-specific. I observe it on several platforms. This patch seems to correct the problem by checking for a zero-length

[openssl.org #1591] get_session_cb callback invoked with no previous session in 0.9.8f

2007-10-17 Thread Stephen Henson via RT
The code was changed when TLS ticket support was added. In that case a zero length session ID can result in a resumed session based on the ticket. It didn't catch the case where ticket resumtion failed and the session legth was zero. This patch should fix it:

Re: [openssl.org #1591] get_session_cb callback invoked with no previous session in 0.9.8f

2007-10-17 Thread [EMAIL PROTECTED] via RT
Yes - the patch at http://cvs.openssl.org/chngview?cn=16691 corrects the problem. Tested with Apache 2.2.6 on Windows and Debian 4.0. -tom- Stephen Henson via RT wrote: The code was changed when TLS ticket support was added. In that case a zero length session ID can result in a resumed