[01:35:04] --- Brandon Allbery has left [01:35:24] --- Brandon Allbery has become available [04:34:20] --- Marc Dionne has become available [05:12:20] --- dev-zero@jabber.org has become available [05:37:20] --- CUDave has left: Disconnected [05:48:04] --- Stephan Wiesand has become available [06:09:44] --- meffie has become available [06:17:32] --- meffie has left [06:19:43] --- meffie has become available [07:42:55] --- asedeno has become available [07:43:23] --- asedeno has left [07:43:39] --- asedeno has become available [07:44:03] --- asedeno has left [07:45:25] --- asedeno has become available [07:45:29] --- asedeno has left [07:52:25] --- asedeno has become available [07:52:35] --- asedeno has left [07:52:51] --- asedeno has become available [07:52:59] --- asedeno has left [07:55:36] --- asedeno has become available [08:13:15] --- deason/gmail has become available [08:20:02] --- Brandon Allbery has left [08:20:11] --- Marc Dionne has left [08:20:53] --- Stephan Wiesand has left [08:21:14] --- Marc Dionne has become available [08:22:00] --- Brandon Allbery has become available [08:24:00] --- Stephan Wiesand has become available [08:47:55] --- Stephan Wiesand has left [09:07:33] --- CUDave has become available [09:43:04] --- meffie has left [09:43:04] --- mmeffie has become available [10:06:23] --- dev-zero@jabber.org has left [11:42:14] --- mmeffie has left [11:42:14] --- mmeffie has become available [11:53:32] --- mmeffie has left [13:27:29] --- dev-zero@jabber.org has become available [13:50:28] --- dev-zero@jabber.org has left [14:07:17] The panic(msg) call at the end of the osi_Panic ifdefs in rx_kcommon causes build trouble for me on recent freebsd. For ukernel, panic is a macro, but for the real kernel build clang complains about a non-literal format string and errors the build. The normal fix is to call panic("%s", msg), but the userland macro is incompatible with that. Is it better to turn userland's thing into a first-rate varargs function, or add more conditionals to rx_kcommon? [14:11:17] osi_Panic should be a varargs function everywhere bar AIX [14:12:37] Is that relevant? [14:14:06] Ah, it's ukernel's panic(), not osi_Panic that's casuing you trouble? [14:14:19] --- Brandon Allbery has left: Lost connection [14:14:30] Right. [14:14:51] I can't see any problem with making ukernel's panic() be more sane. [14:15:13] Okay. [14:17:31] Just watch what you call it - ubik.c also defines panic() [14:18:34] static_inline might be good enough ... ubik should never be built with -DUKERNEL, right? [14:18:52] --- Brandon Allbery has become available [14:19:16] static_inline should be good enough. None of the ubik server bits are used in ukernel, but the client side shims are. [14:22:13] While I've got you here, Simon: the rxgk document has a few sample error codes for what CombineTokens might return in its errorcode field. [14:22:55] (OPCODE, BADETYPE, BADLEVEL, EXPIRED) I guess those are probably all fine as rx aborts... [14:23:10] ... but neither do I need to call them out explicitly, I think. [14:23:12] It really depends on what the caller is trying to do. [14:24:05] If they're going to run a negotiation, then there is a risk. For example, if OPCODE being returned means that we give up on rxgk, and try rxkad instead, or if BADETYPE means that we give up on the current encryption type and try a weaker one, or so on. [14:24:44] But I don't think that the rxgk CombineTokens is well enough defined to use as a negotiation mechanism. So I'm keen to just say "don't do this", and that any error from CombineTokens should fail the attempt at connection establishment. [14:25:28] Yeah, the client should be throwing in all enctypes it is willing to use the first time around, and not saving some in reserve. [14:26:19] I think I'll just remove the errorcode bits from CombineTokens entirely. (Negative diffstats are good, right? ;) ) [14:26:28] That sounds fine to me. [14:53:25] Hmm, AFSCombineTokens doesn't really need the errorcode, either, since it indicates "rxgk not supported" as a zero-length token. [15:09:51] Though ... maybe I do still want an errorcode for AFSCombineTokens. [15:10:03] I guess we can talk about that more, later. [15:51:02] --- Brandon Allbery has left [15:51:36] --- Brandon Allbery has become available [19:05:32] --- deason/gmail has left [20:13:35] --- Brandon Allbery has left [20:15:23] --- Brandon Allbery has become available [21:28:20] --- Brandon Allbery has left [21:28:31] --- Brandon Allbery has become available [21:33:29] --- dev-zero@jabber.org has become available [21:34:26] --- dev-zero@jabber.org has left [21:34:27] --- dev-zero@jabber.org has become available [21:45:26] --- Brandon Allbery has left [21:46:17] --- Brandon Allbery has become available [21:46:51] --- Brandon Allbery has left [21:47:24] --- Brandon Allbery has become available [22:13:56] --- Brandon Allbery has left [22:14:55] --- Brandon Allbery has become available [23:34:34] --- dev-zero@jabber.org has left [23:48:56] --- dev-zero@jabber.org has become available