[00:16:44] --- Russ has left: Disconnected [00:31:38] --- Simon Wilkinson has left [01:27:41] --- dev-zero@jabber.org has become available [01:51:31] --- brantgurga has left [03:06:29] --- cclausen has left [04:13:58] --- brantgurga has become available [04:18:46] --- brantgurga has left [04:58:29] --- brantgurga has become available [05:35:32] --- brantgurga has left [05:37:51] --- Jeffrey Altman has left: Replaced by new connection [05:57:10] --- sxw has become available [06:19:27] --- reuteras has left [06:24:15] --- cclausen has become available [06:49:00] --- haba has become available [07:09:32] --- mmeffie has become available [07:10:27] --- deason has become available [07:52:55] --- haba has left [07:53:13] --- haba has become available [08:08:43] --- Jeffrey Altman has become available [08:11:54] --- mmeffie has left [08:15:55] --- abo has left [08:16:51] --- abo has become available [09:20:33] --- Russ has become available [09:27:04] --- haba has left [09:34:53] --- dev-zero@jabber.org has left [10:38:19] --- dev-zero@jabber.org has become available [11:01:28] --- dev-zero@jabber.org has left: Replaced by new connection [11:01:29] --- dev-zero@jabber.org has become available [11:01:36] --- dev-zero@jabber.org has left: offline [12:58:24] --- sxw has left [14:08:46] --- dev-zero@jabber.org has become available [14:31:48] --- dev-zero@jabber.org has left [14:48:38] --- dev-zero@jabber.org has become available [15:56:44] --- deason has left [15:57:13] --- deason has become available [16:05:15] --- dev-zero@jabber.org has left [16:07:47] --- cclausen has left [17:00:12] --- matt has become available [17:04:36] I've been thinking for a while about some issues I've had with platform mappings for AFS_UCRED. [17:04:47] I think that since it's intended purpose is always be a preprocessor alias to some other type--which may be a platform credential type, or an AFS osi-defined type--it should by definition be a single C identifier. [17:04:58] That is, it should in fact be "typedef AFS_UCRED" not a token sequence "struct AFS_UCRED" in any part of the AFS source. [17:05:12] So I would be proposing to introduce the opaque struct, the typedef from it to AFS_UCRED, and transforming instances of "struct AFS_UCRED..." to "AFS_UCRED..." in source, with some verification work to ensure no bad results in the different clients. [17:05:21] Thoughts on this? [17:08:31] In general, I would prefer that if something is a struct, it be referred to in source code as struct . [17:09:08] One of the things I find extremely annoying about the MIT Kerberos API is that they make everything opaque data types, some of which are simple types, some of which are structs, and some of which are pointers, and you have absolutely no idea what is what. [17:09:16] what happens if it's a struct on some platforms, not on others? [17:09:25] Which means you never know if you're supposed to initialize something to NULL, 0, memset it, or what. [17:09:29] (in general we don't go inside them anyway, and on macos it's an opaque type) [17:09:32] I don't disagree, but I think AFS_UCRED is special, as it is often a mapping to platform opaque types. I am trying to deal nicely with this. [17:10:31] Okay, that's fair. If it's like va_list where the basic type can be completely different on different platforms, that seems reasonable to me. [17:11:01] Is using typedef rather than #define going to give you compiler warnings about type mismatches? [17:11:48] Actually it's going to remove them. You'll then "#define AFS_UCRED (what you need)" and it will always work. [17:12:39] Right now we do just that, but the object macro is only correct if the platform type is introduced with the struct keyword. Not good. [17:15:54] Oh, I see, I misunderstood what your proposal was. I thought you were talking about defining AFS_UCRED with typedef. Yes, that sounds fine. [17:27:28] --- matt has left [21:53:14] --- deason has left [22:59:36] --- dev-zero@jabber.org has become available [23:03:03] --- reuteras has become available [23:26:10] --- haba has become available [23:27:49] Got an openafs-info mail today which has been on grand between 2009-06-11 and 2009-09-02.