Nicolas Baradakis <nbk@sitadelle.com> wrote:
I don't know if it's too soon to plan a release date for 1.1.1, as 1.1.0 has some annoying known problems:
We can put out 1.1.1 as soon as we have fixes for 1.1.0.
- an error on x86_64 machines when reading dictionnaries. - compilation fails on FreeBSD because of rlm_otp. - rlm_eap_peap and rlm_eap_ttls have libtool problems on Debian.
Those 3 alone should result in a 1.1.0.
- pull eap fixes from CVS head to address Debian's issues. I found many changes not in branch 1.1 but I've no idea whether they are safe or not: - moved TLS code from rlm_eap_tls to libeap.
Yes.
- removed record_* functions from global visibility.
Not necessary.
- remove the EAP submodule locking.
Sure, but not necessary.
- change autoconf tests for OpenSSL. - put the eap sessions into a tree, so that looking them up is very fast, and no longer O(n) in the number of sessions.
Not really necessary, but sure. There should only be 10's of ongoing EAP sessions at a time, so it's not critical.
- if the callback fails, do eap_fail() to get an EAP-Failure message and return reject. - use new hex2bin function to be more forgiving for NT-Passwords - make "use_tunneled_reply" work properly for PEAP
Yes.
- fix bug #314 about rlm_otp: the module could have more tests in configure.in so it is silently disabled instead of giving compilation errors.
Yes.
- net-snmp patches. (and move #includes from radius_snmp.h to smux.h)
Yes.
- maybe undelete the database schema for MSSQL server? We could put the file under src/modules/rlm_sql/drivers/rlm_sql_unixodbc for example.
Put the schemas under /doc, or create a /doc/examples and put them there.
- maybe back-port the changes in the makefiles in CVS head? It would fix some warnings and make the autoconf/libtool upgrade easier in the future.
Sure. Alan DeKok.