[00:08:46] --- asedeno has become available [00:09:03] --- andersk has become available [00:11:35] --- kaduk@mit.edu/barnowl has become available [03:52:01] kaduk: Yes, on the line below, CRULE_OPT is built from those three individual variables, according to which architectures are requested by the ARCHFLAGS variable. [05:08:24] --- Simon Wilkinson has left [08:59:31] pre7 on linux with memcache seems nice and speedy [09:16:30] --- Simon has become available [09:18:37] --- haba has left [09:20:50] --- Simon has left [09:40:03] --- haba has become available [09:55:47] --- Russ has become available [10:00:55] Simon: Ah, that will teach me to just look at grep output without context :) [10:11:54] --- jaltman/FrogsLeap has left: Disconnected [10:12:11] --- jaltman/FrogsLeap has become available [10:18:16] heh [11:15:06] nfs_clvnops.c: static int nfs_advlock(struct vop_advlock_args *ap) { struct vnode *vp = ap->a_vp; struct ucred *cred; struct nfsnode *np = VTONFS(ap->a_vp); struct proc *p = (struct proc *)ap->a_id; struct thread *td = curthread; /* XXX */ [11:16:45] Though it looks like he is checking a_flags & F_POSIX before using p [11:19:04] --- haba has left [11:19:44] well, that's one way to deal [11:21:58] Though, looking more carefully, maybe p and td are only used for getting a cred; a_id is passed on unmodified to some other routines. Will look more later. [11:55:22] --- sxw has become available [11:55:50] --- sxw has left [11:57:47] --- haba has become available [14:04:22] fs/nfsclient/nfs_clstate.c:nfscl_lockt uses a_id in a call to nfscl_filllockowner(id, own, flags); but only ever sets l_pid to (pid_t)0. The 'own' value is given as an argument to nfscl_localconflict(). [15:29:36] 1.6.0pre7 is released, right? Should there be a Git tag? [15:30:09] should be, but I don't think it's pushed yet [15:31:17] --- haba has left [15:32:04] It is in a quantum superposition of "released" and "not released", for values of the latter synonymous with "waiting to see if anything is horribly broken". [16:27:57] Hmm, it would be good to check that 5040 builds on darwin. [16:38:50] (why am i seeing some of these in gmail/g+ chat?) [18:01:03] --- andersk has left: Lost connection [18:01:03] --- asedeno has left: Lost connection [18:01:03] --- kaduk@mit.edu/barnowl has left: Lost connection [18:27:43] --- asedeno has become available [18:43:55] --- kaduk@mit.edu/barnowl has become available [19:49:55] have a pre7 tag [20:25:25] --- jaltman/FrogsLeap has left: Replaced by new connection [20:25:26] --- jaltman/FrogsLeap has become available [20:33:53] --- jaltman/FrogsLeap has left: Disconnected [20:38:03] --- jaltman/FrogsLeap has become available [21:22:43] So, should I leave the DARWIN CRULE_FOO_arch as-is, and add something to CRULE_FOO to eat the extra argument? [21:24:28] what extra argument needs to be eaten? (probably) [21:25:10] path/to/foo.c that is in the rule for all the object files foo.o: path/to/foo.c $(CRULE_OPT) path/to/foo.c [21:26:26] lemme try something. [21:26:54] (bsd.kmod.mk is the next commit on top of this one, b.t.w.) [21:27:06] sure [21:36:03] i may have a dead simple fix for you. hang on. i am testing a second dead simple fix on something else right now. i will see after... [21:39:25] sure. simple. instead of molesting the _arch rules, just add a ; echo inside the ) in the main rule, and it eats the extra arg. all done. [21:40:40] Yeah, that's what it looked like once I broke down and actually read that tangled mess in "the line below". [21:42:03] ed moy contributed something like that and i just did it. the problem is unlike the rest of the code where gcc can just take multiple -arch flags, for kernel code you need extra flags. it should be rewritten but each other option sucks differently [21:47:01] Hmm, I suppose I should leave the trailing whitespace in, then, too. [21:53:14] Really put the echo inside the ')' and not outside? [22:07:31] CRULE_OPT=$(shell echo "${ARCH_ppc}" | grep -q -w yes && echo "${CRULE_OPT_ppc} && "; echo "${ARCH_x86}" | grep -q -w yes && echo "${CRULE_OPT_x86} && "; echo "${ARCH_amd64}" | grep -q -w yes && echo "${CRULE_OPT_amd64} && ";echo "lipo -create ${OBJ_ppc} ${OBJ_x86} ${OBJ_amd64} -output $@ ; echo ") [22:08:45] Oh, oops. [22:09:09] which yields a command which ends e.g.: && lipo -create afs_analyze.o.x86 afs_analyze.o.amd64 -output afs_analyze.o ; echo /Users/shadow/Source/openafs/src/afs/afs_analyze.c [22:09:46] you can probably do better. in fact, don't use echo. use true [22:10:57] Inside the quotes would make more sense, yes. [22:11:17] if you use true, no extraneous output [22:13:51] Yeah, yeah. I pushed another patch, you could re-test it ... [22:25:57] --- Russ has left: Disconnected [23:34:50] --- reuteras has become available