[00:09:56] --- Born Fool has left [01:23:07] --- summatusmentis has left [02:33:39] --- Simon Wilkinson has become available [02:43:35] --- Russ has left: Disconnected [02:50:48] --- Simon Wilkinson has left [02:50:57] --- Simon Wilkinson has become available [02:54:37] --- kaj has become available [02:58:18] --- Simon Wilkinson has left [03:48:25] --- jaltman has left: Replaced by new connection [03:48:26] --- jaltman has become available [04:07:34] --- kaj has left [06:16:32] --- jaltman has left: Disconnected [06:16:54] --- jaltman has become available [08:21:58] --- kaj has become available [09:22:37] --- mdionne has become available [09:29:50] --- tharidufernando has become available [09:47:32] --- mdionne has left [10:06:15] --- mdionne has become available [10:12:47] --- kaj has left [10:24:01] --- deason has become available [10:48:47] --- tharidufernando has left [10:57:58] --- summatusmentis has become available [11:21:00] --- Simon Wilkinson has become available [11:27:20] --- Russ has become available [11:40:39] --- Simon Wilkinson has left [12:16:48] --- Jeffrey Altman has become available [12:17:10] --- jaltman has left: Replaced by new connection [12:17:10] --- jaltman has become available [13:04:10] --- kaj has become available [14:53:52] --- jaltman has left: Disconnected [16:25:25] --- kaj has left [16:25:30] --- kaj has become available [17:59:29] For kicks, I tried not passing MNTK_MPSAFE in vfs_mount(), which went ... poorly. A lock order reversal between the glock and Giant (aka big kernel lock) leaves me unable to do anything with my keyboard. [18:20:25] Regarding my aklog coredump, I do have a working valgrind now; it's claiming lots of invalid writes in rx_packet.c, with some in rx.c and elsewhere. It also fails to actually get tokens, and aklog prints the "a pioctl failed while getting tokens" message, so it seems like valgrind is getting confused by our syscall. [18:26:11] You get oodles and oodles of valgrind errors from savecontext/setcontext. I tried to dig out the errors that didn't reference that stuff when I was debugging before. [18:26:32] I'm not sure if those indicate actual bugs in our context stuff, or if it's just valgrind being confused. [18:26:36] It may be that they're real bugs. [18:28:31] Yeah, I don't think I'm actually getting the failing codepath when running in valgrind, at the moment. I may need to recompile valgrind with a tweaked header file. [18:40:34] I wonder if it's possible to build aklog with pthreads. [18:40:52] I think Derrick (?) said he was going to try. [18:41:25] Eventually we should be building everything that way so that we can throw out our own threading library, which causes tons of headaches. [18:41:38] Ubik work has to be finished first, though. [18:57:32] Hm, reading valgrind's README_MISSING_SYSCALL_OR_IOCTL would make it seem that I need to write some actual code to wrap the AFS syscall/ioctl. But presumably this has already been done, somewhere ... [20:15:03] --- mdionne has left [20:49:38] --- jaltman has become available [21:59:59] --- deason has left [22:03:26] --- reuteras has become available [22:58:05] --- kaj has left [23:06:03] --- haba has become available [23:53:28] --- kaj has become available [23:59:35] --- kaj has left [23:59:49] --- kaj has become available