Timo Sirainen skrev:
On Mon, 2007-07-02 at 18:08 +0200, Peter Eriksson wrote:
+ AC_CHECK_LIB(gss, __gss_userok, [
+ AC_DEFINE(HAVE___GSS_USEROK,,
+ Define if you have
__gss_userok())
+ ])
Instead of changing $LIBS, I think you can use:
AC_CHECK_LIB(gss, __gss_userok, [ ...
],, `krb5-config --libs gssapi`)
Ok, I'll try that instead (I used the changing stuff since similar stuff
was done earlier (with CFLAGS) :-)
Btw. I think that the Heimdal implementation of GSSAPI has a similar
function but called gss_userok() (without the leading two "_"
characters) that does something similar - and users of Heimdal around
that can verify or deny that?
+#if 0 /* Added 070702 Peter Eriksson <[EMAIL PROTECTED]> - ref cyrus-sasl
Solaris 8/9 has problems with NO_OID */
+ GSS_C_NT_USER_NAME,
+#else
GSS_C_NO_OID,
+#endif
Is this a common problem? Should it be a setting? Although with your
change this code is never even executed, right?
Dunno actually - When I was reading the Cyrus-SASL GSSAPI-code I saw
that they have a configure-check for
GSS_C_NT_USER_NAME in their code base and a comment that Solaris 8/9
doesn't like GSS_C_NO_OID
in that call so at first I thought that was the reason things didn't
work - however it wasn't the problem I later figured out...
Anyway, as you can see I disabled that change (and with the
_gcc_userid() code it's not used anyway).
+ name = p_strndup(request->auth_request.pool,
+ (unsigned char *)outbuf.value + 4,
+ outbuf.length - 4);
+
+ if (!name) {
+ auth_request_log_error(&request->auth_request,
"gssapi",
+ "Invalid response size");
+ auth_request_fail(&request->auth_request);
+ return;
+ }
name is never NULL.
Ok, I'll remove those checks.
+ free(name);
And you shouldn't free() it either.
And I'll remove the free() calls too... :-)
I've uploaded new versions of the patches to:
ftp://ftp.ifm.liu.se/pub/unix/dovecot/
I've also put a Solaris 10 SMF manifest there (dovecot.xml).
Btw, is it intentional that the first line in configure.in starts with
an "x"?
- Peter
