Mark Phalan wrote:
> On Fri, 2009-01-16 at 13:38 -0500, Peter Shoults wrote:
>   
>> I am putting this comment in last - ever file needs the copyright 
>> updated to 2009.  Ignore the few below that say copyright.
>>     
>
> Yup. I'll update the year for all CRs.
>
>   
>>>> 6768868 CPPFLAGS from Makefile.mech_krb5 get clobbered for pam_krb5
>>>>         and pam_krb5_migrate
>>>> http://cr.opensolaris.org/~mbp/6768868-wsdiff_pam_krb5/
>>>>     
>>>>         
>> Copyright again.  Also, by putting Makefile.mech_krb5 after 
>> Makefile.pam_modules, you will now lose,
>> CPPFLAGS    += -D_REENTRANT that was defined in Makefile.pam_modules, 
>> right?  Is that ok, or should you
>> add the D_REENTRANT to the Makefile.mech_krb5?
>>     
>
> As Makefile.mech_krb5 only adds (+=) to CPPFLAGS any previously set
> CPPFLAGS will still be set including -D_REENTRANT.
> i.e this is ok.
>
>   
>>     
>>>> 6776724 krb5_preauth_request_context_init leaks memory
>>>> http://cr.opensolaris.org/~mbp/6776724-PKINIT_memleaks/
>>>>     
>>>>         
>> So, for long-lived apps, are they going to keep re-opening and init'ing 
>> the plugin info, which will keep growing memory usage?
>>     
>
> The plugin info will be re-inited but it won't leak. OpenSSL is smart
> enough not to leak memory in this case. It will initialize the global
> data once. I've also verified that this is the case.
>
>
>   
>>   If so, that 
>> seems like another memory leak.
>>     
>
> See above. No new memory leak is introduced.
>
>
>   
>>   Copyright.
>>     
>>>> 6777148 idmap fails to auto-discover AD due to ldap_sasl_bind failure
>>>> http://cr.opensolaris.org/~mbp/6777148_init_creds_realm/
>>>>     
>>>>         
>> locate_kdc.c should call and check the results of the call to 
>> krb5_is_referral_realm in the if statement using the same format as do 
>> the other two files.  Notably, locate_kdc.c uses a test of != TRUE, 
>> which the others do not.
>>     
>
> The test in locate_kdc.c is a negative test for
> krb5_is_referral_realm(). If you'd like I can change it to:
>
> if (code && !krb5_is_referral_realm(realm)) {
>
> I don't mind either way.
>   

Yeah - I think that is more uniform for all three files/routines.
>   
>>>> 6782682 krb5_recvauth() should return NULL for auth_context on failure
>>>> http://cr.opensolaris.org/~mbp/6782682-krb5_recv_auth_null/
>>>>
>>>>     
>>>>         
>> Hum - looking at this code, I do not see why this is right:
>>
>> 240 cleanup:;
>>  241     if (retval) {
>>  242         if (local_authcon) {
>>  243             krb5_auth_con_free(context, *auth_context);
>>  244             /* Solaris Kerberos */
>>  245             *auth_context = NULL;
>>  246         } else if (local_rcache && rcache != NULL) {
>>  247             /* Solaris Kerberos */
>>  248             (void) krb5_rc_close(context, rcache);
>>  249             krb5_auth_con_setrcache(context, *auth_context, NULL);
>>  250         }
>>  251     }
>>
>> I do not see why if we have an error, we could not have both 
>> local_authcon set as well as having a local_rcache that needs to get 
>> cleaned up.  
>>     
>
> At first glance you seem to be correct. However thats not this bug.
>   

Is that a different bug?  If so, perhaps we should open on to track it.
>   
>> It looks like if we have local_authcon, we will never test 
>> and clean up the local_rcache if needed.
>>
>>
>>     
>>>> 6784485 keys for kadmind princs should be created with all supported
>>>>         enc-types
>>>> http://cr.opensolaris.org/~mbp/6784485-enctypes_for_default_kadmin_princs/
>>>>     
>>>>         
>> I am running out of time and have to leave soon.  I will review this 
>> last file on Monday if that is ok, or someone else can do the last one.
>>     
>
> No problem. Thanks for the review so far!
>
> -M
>
> _______________________________________________
> kerberos-discuss mailing list
> kerberos-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/kerberos-discuss
>   


Reply via email to