[02:25:51] Other users of our com_err were broken when we renamed all of our symbols. [04:50:08] > Other users of our com_err were broken when we renamed all of our symbols. [04:50:16] we provided a header to fix that [04:57:25] Only if you build with AFS_OLD_COM_ERR, IIRC [04:58:27] yes. but that's not "broken" per se. we advertised it [07:39:00] --- deason has become available [08:20:36] So we can advertise renaming the library too? I just think it's wrong of us to be building, and shipping, a libcom_err which has none of the symbols you'd expect to find in such a library. [08:54:23] there were objections at the time to renaming the library. I don't remember what they were. [08:55:45] The problem is that if you're doing -lkrb5 -lcom_err , you _really_ don't want our com_err. [08:57:09] rmember the history. those that are using the afs com_err really do want the afs com_err. [08:57:53] and they wanted the afs com_err to be like normal com_err because there should only be one com_err. [08:58:50] at this point its been three years since we changed to AFS_OLD_COM_ERR. renaming the library is probably ok. We sent the warning that was necessary. However, I'm sure that folks will still scream [09:00:11] You broke it once, and now you're going to break it again. So yes, I'll scream. [09:01:04] but, GONE [09:04:12] It's not true that those that are using the afs com_err really do want it. If I'm linking a third-party Kerberos library, I really don't want it. It's not what I was built against, and linking will fail. [09:07:12] --- rra has become available [09:08:33] I wonder if we should have an autoconf check, and use a system com_err if we can find one, and only fall back to afs_com_err if we can't. [09:09:04] What does Debian do? Because our com_err bundling must be against Debian policy... [09:12:28] --- jaltman/FrogsLeap has left: Disconnected [09:26:54] With AFS, we don't do anything special, just whatever the code does by default. [09:27:05] Which I think ends up with bad error messages for the stuff that mingles Kerberos and AFS. [09:27:32] Embedding our own copy of com_err isn't great, but on the other hand it's substantially forked from everyone else's, so it's not an obvious example of just a copy of the code. [09:56:28] --- pod has become available [11:42:22] --- jaltman has become available [11:43:10] Nice. Mac OS X breaks ABI compatibility for GSSAPI by using #pragma pack(push, 2) [11:44:16] is that MIT GSSAPI or something else? [12:04:44] So, if you build against Apple's gssapi.h header than you get different structure packing than if you build against anyone elses. [12:05:28] So you can no longer dynamicaly link in a different GSSAPI library (be it MIT, Heimdal, GSI, ...). Same goes in the other direction - you can't build against an MIT header, and dynamically link in the Apple library. [12:06:24] So Apple is shipping a header which is different than the one that MIT provided to them as part of KfM? that is what I'm attempting to determine [12:07:00] Well, I don't know what MIT gave them. But it's a different header from the one that MIT ship as part of their Kerberos releases. [12:07:08] If they are using the header that MIT provided as part of KfM, then it isn't Apple's fault. [12:08:19] I no longer have access to the KfM repo so I can't check. You might want to send lxs an e-mail and ask her. You might also point it out to lha and see if it can be fixed for Lion [12:09:03] I suspect it was a KfM-ism hold over from the earlier pre-Mach Mac releases [12:09:26] --- jaltman has left: Disconnected [12:09:31] Actually, I tell a lie. The offending #pragma pack() is in the 1.8.x release [12:09:58] So, it's MIT's fault. [12:16:13] http://krbdev.mit.edu/rt/Ticket/Display.html?user=guest&pass=guest&id=5643 [12:17:35] It's the release why GSSAPI segfaults Firefox 4 and Thunderbird on 64 bit Mac OS X [12:17:41] s/release/reason/ [12:18:05] Thankfully David Bienvenu tracked down the release before they just disabled GSSAPI. [12:19:14] related: http://anonsvn.mit.edu/viewvc/krb5/trunk/src/lib/gssapi/generic/gssapi.hin?r1=18396&r2=19871&pathrev=20019 [12:23:08] oof [12:25:20] I've sent email to krbdev. We'll see what the response is, but I won't hold my breath. [12:25:28] I wonder if it's worth also opening a radar. [14:00:50] --- deason has left [14:00:51] --- deason has become available [14:11:26] --- jaltman has become available [14:14:10] I would open a radar as well [14:14:40] --- jaltman has left: Disconnected [14:14:47] --- jaltman has become available [15:09:11] --- jaltman has left: Disconnected [15:09:23] --- jaltman has become available [15:20:37] --- jaltman has left: Disconnected [15:20:47] --- jaltman has become available [15:21:51] --- jaltman has left: Disconnected [15:36:12] --- jaltman has become available [15:47:07] --- jaltman has left: Disconnected [15:57:41] --- jaltman has become available [15:57:43] --- jaltman has left: Disconnected [16:01:22] --- jaltman has become available [16:04:11] --- jaltman has left: Disconnected [16:29:59] --- jaltman has become available [16:31:49] --- jaltman has left: Disconnected [16:46:42] --- deason has left [17:32:06] --- rra has left: Disconnected [17:52:24] --- Russ has become available [18:08:07] --- kula has left [18:09:40] --- kula has become available [19:08:35] --- deason has become available [20:14:12] --- Russ has left: Disconnected [21:48:03] --- jaltman has become available [21:48:18] --- jaltman has left: Disconnected [22:15:59] --- jaltman has become available [22:16:10] --- jaltman has left: Disconnected [22:16:16] --- jaltman has become available [22:29:40] --- deason has left [23:25:16] --- Simon Wilkinson has left