[00:13:34] --- jaltman/FrogsLeap has left: Replaced by new connection [00:13:35] --- jaltman/FrogsLeap has become available [00:15:50] --- jaltman/FrogsLeap has left: Replaced by new connection [00:15:51] --- jaltman/FrogsLeap has become available [00:56:14] --- lars.malinowsky has become available [01:46:05] --- abo has become available [02:17:32] --- Russ has left: Disconnected [04:27:15] --- lars.malinowsky has left [04:27:27] --- lars.malinowsky has become available [04:39:47] --- jaltman/FrogsLeap has left: Disconnected [04:40:51] --- jaltman/FrogsLeap has become available [06:12:55] --- reuteras has left [06:16:43] --- jaltman/FrogsLeap has left: Disconnected [06:17:38] --- jaltman/FrogsLeap has become available [06:51:57] --- matt has become available [07:00:01] --- jaltman/FrogsLeap has left: Disconnected [07:28:20] --- deason has become available [07:28:51] --- lars.malinowsky has left [08:22:17] --- jaltman/FrogsLeap has become available [08:54:46] --- rra has become available [10:30:51] so, how much do i want to break down my netbsd cache manager port for submission? [10:31:22] can i see a diff of the whole thing? just a diff in afs? [10:34:01] mostly in src/afs, but not entierly [10:34:53] no, i mean, drop the diff in afs [10:38:38] http://poghril.kollasch.net/openafs-netbsd-20110503-00.txt [10:39:28] i can push that to a public cell if you really need it in afs [10:40:43] get rid of dumptv() [10:40:50] and the commented out calls of it [10:41:05] and the commented out timo printf. [10:41:22] yeah, i planed on removing most all of that kind of thing [10:41:38] i'm more concerned about the stuff i did in src/external [10:41:44] the if 0'd code in TryEvictVCache [10:41:48] yeah, getting there [10:42:21] just make the if 1'd code the default in afs_root and nuke the old code [10:42:56] i'm still not quite sure if either one is right :-) [10:43:59] Do we really prefer [10:44:01] - memset(avc, 0, sizeof(struct vcache)); + memset(avc, 0, sizeof(*avc)); [10:44:07] ? [10:44:15] er, "is probably right" [10:44:56] matt: i may have been following the style guide of other projects i work on more closely than i should have [10:45:09] I think go with type explicit, for this. [10:45:20] why can aes.c not include string.h? [10:45:51] basically, the external changes are probably all not ok as is, ignoring how they get pushed. [10:46:04] In osi_vcache.c are you renaming to afs_global_lock because it will not be a mutex? [10:46:10] yes, the external stuff cannot be pushed directly to us; you need to give it to heimdal and then we pull it [10:46:12] submit all but those as one patch [10:46:13] It seems to be now. [10:46:14] but what's there doesn't look right anyway [10:46:42] Also, there's a c++ comment that looks suspicous later #ifdef AFS_NBSD50_ENV - mutex_init(&avc->rwlock, MUTEX_DEFAULT, IPL_NONE); + //rw_init(&avc->rwlock); #else [10:46:43] yeah. 1) not for us (though we can help fix upstream) but 2) what's there is wrong, probably if it should be done, it should be done some other way [10:47:18] > memset for new code, we go either way, but if that's the only change you're making, leave it alone [10:47:47] clean up the commented out/if 0'd stuff, drop the external and push the rest as one patch, and we'll go over that. once it's done, we can figure out what the real issues in external are. reasonable? [10:48:00] deason: bzero is on the list for eventual removal from the netbsd kernel [10:48:25] not that we can't solve that case with a define ... [10:48:40] so just change it to a memset and don't otherwise rewrite the arguments [10:48:46] right [10:48:54] jakllsch: I'm referring to changing the third memset argument from sizeof(type) to sizeof(*var) [10:49:02] we already made our code use mem* a while ago, and #define back to b* where needed. [10:49:22] ah [10:49:33] I just dont think sizeof(*var) is the best idea. The code should know what the type is, and say so. [10:50:00] k. [10:50:09] the external/heimdal stuff with headers seems like it's fixable with buildsys changes; I'm a little confused by the "#define MD5_Init MD5Init" stuff, though [10:50:12] the stuff in external i wasn't planing on pushing anytime soon [10:50:14] There are cases (not this) where a compiler will do the wrong thing with separate compilation. [10:50:23] matt: then you change the var type and you don't even get a warning when the 3rd memset arg is wrong [10:50:46] It's worse to get the wrong result. [10:51:05] but that argument doesn't matter for now imo; we have both ways in the tree, I'd just prefer not to change those intermingled with netbsd portability issues [10:51:21] Sure, that's fine. [10:52:35] huh, that aes.c change i don't even need ... i didn't need to compile that file at all ... [10:53:02] deason: huge cpp mess [10:53:25] the change to afs_pd_putString in afs_pioctl.c should also be separate... (I'd also wonder what we're doing that even makes that noticeable) [10:53:57] jakllsch: what's the problem with renaming those symbols to hc_MD5_Init like heimdal usually does? [10:55:47] deason: src/crypto/hcrypto/kernel/config.h:48:1: error: malloc redefined [10:56:13] basically, i found it easier to just use the ones already in the kernel [10:56:46] i should note, my build system is external to the normal tree and build process [10:58:21] but in general, everything that's nbsd-only code can probably go in all at once; try to pull out things that hit other platforms (like afs_pd_putString or changing AFS_MAXIOVCNT) [11:00:50] --- Simon has become available [11:01:11] matt: the afs_global_mtx -> afs_global_lock change was mostly to mirror Solaris and Irix [11:01:35] What does freebsd do? [11:02:06] afs_global_mtx it looks like [11:02:43] I think we should keep the 4bsd ports as consistent as we can. [11:02:57] yes, I think you'll find nbsd more similar to the fbsd cm than solaris [11:03:25] some parts yes, other parts no [11:04:05] but whatever, it doesn't really matter much to me what the mutex is called [11:04:48] I think you'll find agreement with that, which is why not changing it may seem easier [11:16:08] --- Simon has left [12:32:46] --- ksumner has become available [12:47:32] --- jaltman/FrogsLeap has left: Replaced by new connection [12:47:33] --- jaltman/FrogsLeap has become available [12:59:05] --- phalenor has left [12:59:11] --- phalenor has become available [15:08:49] --- meffie has become available [15:31:10] --- ksumner has left [15:39:44] --- deason has left [15:48:15] --- Simon Wilkinson has become available [15:55:45] --- matt has left [17:28:36] --- matt has become available [18:16:16] --- rra has left: Disconnected [18:34:51] --- Russ has become available [19:13:20] --- meffie has left [23:47:33] --- reuteras has become available