From hartmans@MIT.EDU Sat Nov 23 00:22:39 1996 Received: from MIT.EDU (SOUTH-STATION-ANNEX.MIT.EDU [18.72.1.2]) by rt-11.MIT.EDU (8.7.5/8.7.3) with SMTP id AAA15681 for ; Sat, 23 Nov 1996 00:22:39 -0500 Received: from STARKILLER.MIT.EDU by MIT.EDU with SMTP id AA17846; Sat, 23 Nov 96 00:22:38 EST Received: by starkiller.MIT.EDU (5.x/4.7) id AA03015; Sat, 23 Nov 1996 00:22:37 -0500 Message-Id: <9611230522.AA03015@starkiller.MIT.EDU> Date: Sat, 23 Nov 1996 00:22:37 -0500 From: hartmans@MIT.EDU Reply-To: hartmans@MIT.EDU To: krb5-bugs@MIT.EDU Subject: make check fails under NetBSD X-Send-Pr-Version: 3.99 >Number: 232 >Category: krb5-admin >Synopsis: make check fails under NetBSD >Confidential: no >Severity: serious >Priority: high >Responsible: bjaspan >State: closed >Class: sw-bug >Submitter-Id: unknown >Arrival-Date: Sat Nov 23 00:23:00 EST 1996 >Last-Modified: Mon Nov 25 17:49:46 EST 1996 >Originator: Sam Hartman >Organization: mit >Release: 1.0-development >Environment: System: SunOS starkiller 5.4 Generic_101945-37 sun4m sparc >Description: Other kadmin tets succeed. I will append the fialure data; I think Marc is looking at it. NetBSD fails the gssapi test because kadmin fails to start. The other tests are successful >How-To-Repeat: >Fix: >Audit-Trail: State-Changed-From-To: open-analyzed State-Changed-By: marc State-Changed-When: Sat Nov 23 03:09:32 1996 State-Changed-Why: I instrumented a whole bunch of stuff. I know what the problem is, basically, but no clue *why* it is happening. kadmind sets krb5_defkeyname to the admin keytab name, which is broken, but it it also what we've got for now. If KRB5_KTNAME is set, that gets used instead of the name in the kdc.conf file. the kdc.conf file is correct; the keytab is in tmpdir/admin-keytab and is correct, but KRB5_KTNAME is set most bogusly: (gdb) p ((char **) environ)[21] $46 = 0xf7bfdc9b "KRB5_KTNAME=/u1/var/tmp/krb5/build/kadmin/testing/krb5-test-root/ovsec_adm.srvtab" The actual failure is that the principal isn't in the keytab, which is consistent with the above, and all the work I did in the debugger. At this point, it's 3am, and I don't want to work on this anymore. I'm convinced the problem is in the tests, but I don't have a real clue where. I think that this is relevant: ./kadmin/testing/scripts/env-setup.shin:KRB5_KTNAME=$K5ROOT/ovsec_adm.srvtab; export KRB5_KTNAME but I haven't finished tracking it down. again, it seems really weird that this is only happening on netbsd. I recommend that this not hold up the release, but this is Ted's decision. From: Sam Hartman To: Marc Horowitz Cc: Barry Jaspan , krb5-bugs@MIT.EDU, krbdev@MIT.EDU Subject: Re: krb5-admin/232: make check fails under NetBSD Date: 23 Nov 1996 14:07:55 -0500 While now that we are fairly sure this is a test problem, I agree it probably shouldn't hold up the release, I wanted to point out that it happens on OSF as well as NetBSD. From: Marc Horowitz To: Barry Jaspan , hartmans@MIT.EDU Cc: krbdev@cygnus.com Subject: Re: krb5-admin/232: make check fails under NetBSD Date: Sat, 23 Nov 1996 03:18:48 -0500 (EST) `Marc Horowitz' changed the state to `analyzed'. State-Changed-From-To: open-analyzed State-Changed-By: marc State-Changed-When: Sat Nov 23 03:09:32 1996 State-Changed-Why: I instrumented a whole bunch of stuff. I know what the problem is, basically, but no clue *why* it is happening. kadmind sets krb5_defkeyname to the admin keytab name, which is broken, but it it also what we've got for now. If KRB5_KTNAME is set, that gets used instead of the name in the kdc.conf file. the kdc.conf file is correct; the keytab is in tmpdir/admin-keytab and is correct, but KRB5_KTNAME is set most bogusly: (gdb) p ((char **) environ)[21] $46 = 0xf7bfdc9b "KRB5_KTNAME=/u1/var/tmp/krb5/build/kadmin/testing/krb5-test-root/ovsec_adm.srvtab" The actual failure is that the principal isn't in the keytab, which is consistent with the above, and all the work I did in the debugger. At this point, it's 3am, and I don't want to work on this anymore. I'm convinced the problem is in the tests, but I don't have a real clue where. I think that this is relevant: ./kadmin/testing/scripts/env-setup.shin:KRB5_KTNAME=$K5ROOT/ovsec_adm.srvtab; export KRB5_KTNAME but I haven't finished tracking it down. again, it seems really weird that this is only happening on netbsd. I recommend that this not hold up the release, but this is Ted's decision. From: "Barry Jaspan" To: marc@cygnus.com Cc: hartmans@MIT.EDU, krbdev@cygnus.com Subject: Re: krb5-admin/232: make check fails under NetBSD Date: Sat, 23 Nov 96 15:48:45 -0500 kadmind sets krb5_defkeyname to the admin keytab name, which is broken, but it it also what we've got for now. If KRB5_KTNAME is set, that gets used instead of the name in the kdc.conf file. the kdc.conf file is correct; the keytab is in tmpdir/admin-keytab and is correct, but KRB5_KTNAME is set most bogusly: (gdb) p ((char **) environ)[21] $46 = 0xf7bfdc9b "KRB5_KTNAME=/u1/var/tmp/krb5/build/kadmin/testing/krb5-test-root/ovsec_adm.srvtab" I don't understand. When you run start_servers, the kadmind keytab is created in the test root dir (build/kadmin/testing/kr5-test-root/ovsec_adm.srvtab; I presume in the example about that /u1/var/tmp/krb5/build is your build directory). This path is also the value that is specified in the kdc.conf which is created in the krb5-test-root directory. Additionally, for convenience, the environment variable KRB5_KTNAME is set by env-setup.sh, although I no longer really remember why. The upshot is that KRB5_KTNAME and kdc.conf always agree. You say that the keytab is in tmpdir/admin-keytab. That means you must have mucked with the kdc.conf generated by start_servers by hand, and created the keytab by hand, because otherwise it wouldn't be there. In which case, I'm not surprised that the tests are breaking. Perhaps my interpretation of what you said is not correct. In any case, since I know this code better than anyone, I request that no one make changes to the admin test system to "fix" this problem until I can analyze it on Monday or get more details. Barry State-Changed-From-To: analyzed-closed State-Changed-By: tytso State-Changed-When: Mon Nov 25 14:29:13 1996 State-Changed-Why: Problem fixed. From: "Theodore Y. Ts'o" To: hartmans@MIT.EDU Cc: krb5-bugs@MIT.EDU Subject: Re: krb5-admin/232: make check fails under NetBSD Date: Mon, 25 Nov 1996 14:28:52 -0500 For the record, the problem was due to the KRB5_KTNAME environment variable set before running "make check". I've checked in changes from Ezra which fix this problem, by making the dejagnu tests clear KRB5_KTNAME before starting kadmind. - Ted State-Changed-From-To: closed-closed State-Changed-By: hartmans State-Changed-When: Mon Nov 25 17:46:48 1996 State-Changed-Why: OK, the fix appears to have fixed things. I'm still not sure how the environment variable was getting set; Ted's hypothesis is that I was confused and sourced env-setup.sh; I sort of doubt this a lot, but I am not sure. >Unformatted: