[01:24:23] --- Simon Wilkinson has become available [01:24:45] --- Simon Wilkinson has left [02:51:04] --- abo has left [02:51:04] --- abo has become available [05:28:33] --- Simon Wilkinson has become available [05:29:41] --- Simon Wilkinson has left [05:30:14] --- ballbery has left [05:31:12] --- ballbery has become available [06:06:32] --- mvita has left [06:06:37] --- mvita has become available [07:18:46] --- ballbery has left [07:19:38] --- ballbery has become available [07:41:00] --- ballbery has left [07:49:57] --- ballbery has become available [08:50:40] --- jaltman/FrogsLeap has left: Disconnected [08:53:51] --- jaltman/FrogsLeap has become available [08:57:31] --- jaltman/FrogsLeap has left: Disconnected [09:00:47] --- jaltman/FrogsLeap has become available [09:55:25] --- deason has become available [10:03:46] --- deason has left [10:04:45] --- deason has become available [10:05:35] --- deason has left [10:06:02] --- deason has become available [10:31:47] --- gendalia87003 has left [10:36:47] --- Simon Wilkinson has become available [10:52:07] --- Simon Wilkinson has left [10:52:21] --- Simon Wilkinson has become available [11:05:07] --- Simon Wilkinson has left [11:05:18] --- Simon Wilkinson has become available [11:14:07] --- Simon Wilkinson has left [11:14:37] --- Simon Wilkinson has become available [11:25:58] --- Simon Wilkinson has left [11:26:29] --- Simon Wilkinson has become available [11:29:18] --- Simon Wilkinson has left [11:58:39] --- deason has left [11:58:40] --- deason has become available [12:33:54] Marc: is there a way to conditionalize at the source level whether code is being built for afspag.ko or libafs.ko? My latest build of 10841 seems to be choking while building afspag.ko, claiming that hc_RAND_bytes is undefined. (I'm failing to see where rx.o is getting pulled into the *_PAGOBJS, either.) Note that 10841's patch set 2 still has other build errors, so the fedora builder(s) is the one to look at for this error. [12:34:19] --- deason has left [12:34:20] --- deason has become available [12:57:03] Ben: rx.c gets compiled as rx_pag.o for afspag.ko, so you'll need to add whatever provides the RAND_bytes function to PAGOBJS, which might end up requiring several other files for dependencies [12:57:34] Ah. Yup, it will require a number of other things. Thanks for the clarification. [12:58:11] At least that means I can move the lock initialization back to afs_osi.c. [13:00:43] it's necessary to build separate objects for some source files because the kernel build system will not define a few variables when the same object is part of several modules, causing errors in a few header files [13:01:53] Ah. It's a little confusing that linux is the only platform that actually cares about the *_PAGOBJS, but that smells like a historical artifact. [13:02:30] --- deason has left [13:02:31] --- deason has become available [13:03:00] I also had some fun with building rand-fortuna.c in the kernel because symbols like 'current' and 'u' (which it uses as local variables) are in the global preprocessor namespace in the kernel. Some of the errors were pretty arcane. [13:06:58] --- deason has left [13:06:59] --- deason has become available [13:11:02] yeah we've had name clashes like that in the past where things have been renamed as afs_xxx. a bit trickier with external code. #define u is just evil [13:12:18] But tweaking make_kbuild_makefile.pl to make copies and sed them instead of links smells like a terrible idea. [13:53:55] Is there a rule of thumb for when an object will need to be built twice for the two modules? [13:59:10] --- ballbery has left [13:59:15] --- ballbery has become available [14:09:47] not really, so far i think it's just: split them when there's a known build error. it's needed if the source file ends up pulling a kernel header that uses KBUILD_MODNAME [14:10:39] Okay. I expect that the hcrypto sources will not be pulling in any kernel headers that know about KBUILD_MODNAME :) [14:21:16] --- ballbery has left [14:21:19] --- ballbery has become available [14:21:24] --- mvita has left [14:27:17] those files don't include much if any kernel specific headers, so it's very likely not a problem [14:28:43] (I already sent an updated patch to gerrit, so the buildbot will get it in due time. [14:28:50] ) [15:01:09] --- meffie has left [15:17:04] --- Simon Wilkinson has become available [15:19:36] --- deason has left [15:23:46] --- Simon Wilkinson has left [15:24:36] --- Simon Wilkinson has become available [15:31:33] --- Marc Dionne has left [15:46:51] --- Simon Wilkinson has left [16:08:58] --- Marc Dionne has become available [16:19:20] --- ballbery has left [16:23:52] --- ballbery has become available [20:31:52] Yay I made afspag.ko build, that was entirely too complicated.