Re: Cannot get EAP-TLS to work with FreeRADIUS 0.7

[Aron Silverton]

Artur Hecker wrote:

hi

Exactly.  Which is strange that it did compile with 0.9.6b.  I didn't
capture all of the output, so I'm not sure what, if anything, it said
about the function calls.

For the record, the stable snapshot of OpenSSL 0.9.7 from 2002-10-30
works fine.  Maybe Raghu can update the EAP document?

Aron

for as far as i know, raghu is not currently reading the list... (not
sure about it).

i actually didn't get your point: for the eap/tls module you need the
openssl0.9.7 which is still beta, right? you said that there are some
functions not available in the 0.9.6x version, no? so why would you ask
if there is any reason for using the beta version? did i miss something?

Yes, the EAP document with the FreeRADIUS 0.7 tar ball indicates that OpenSSL 0.9.6b or later (for example, 0.9.6c - 0.9.6g) should work. Others were having problems with that release, so I tried it, verified that it didn't work, tried to find out why it doesn't work (rlm needs more recent function calls), and found a version that definitely does work. I'm not saying I figured something out that nobody else new, heck, I already new it and was using 0.9.7 snapshots on other machines.

A how-to from UMD indicated that as far as 0.9.7 is concerned, a snapshot, as opposed to a stable snapshot, was needed. That doesn't seem to be the case any more as the stable snapshots are all much more recent than the snapshots available at the time that the how-to was written.

and i think that the configure routines for the rlm_eap_tls only test
the init functions of libssl and libcrypto, which work in both release
and betas... it there any version testing in there? i don't know. and
are you sure, the compilation worked at the same point of time as he
built the rlm_eap_tls? he could have built it before and then could have
recompiled with "usual" configure, that is without rlm_eap_tls and
that's the "old" rlm which the nex server tried to load due to the old
config. then, the linker can't resolve symbols, that's all...

In any event, the configure routines should indicate which version is needed, and if it is not found, that should be reported as well. Otherwise, wrt to this module anyway, what's the point?

What I didn't figure out, is why, even if configure succeeds, make did not fail when it tried to compile code referencing functions that did not exists in any of the include files?

ciao
artur

--
Aron J. Silverton
Senior Staff Research Engineer
Motorola Laboratories, Networks and Infrastructure Research
Motorola, Inc.

Telephone: 847-576-8747
Fax: 847-576-3240
mailto: [EMAIL PROTECTED]


- List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html