[00:14:04] --- steven.jenkins has left
[00:16:59] --- steven.jenkins has become available
[00:21:21] --- kaj has become available
[00:36:33] --- deason has become available
[00:36:40] --- deason has left
[00:46:14] --- kaj has left
[02:46:07] --- Jeffrey Altman has left: Replaced by new connection
[02:46:09] --- Jeffrey Altman has become available
[04:23:47] --- kaj has become available
[04:49:46] --- dev-zero@jabber.org has become available
[07:06:18] --- dev-zero@jabber.org has left
[07:06:20] --- dev-zero@jabber.org has become available
[08:10:45] --- pod has left
[10:18:51] --- jaltman has left: Disconnected
[10:19:00] --- jaltman has become available
[10:20:33] --- pod has become available
[10:56:25] --- dev-zero@jabber.org has left
[11:01:05] --- kaj has left
[11:05:37] --- kaj has become available
[11:12:42] --- kaj has left
[11:39:50] <kaduk@mit.edu/barnowl> (still new to git/etc.)
If I have pushed to gerrit from a local branch, that means that I
need to merge instead of rebase when I git pull from remote/master on
that branch?
[11:51:22] <Simon Wilkinson> No
[11:51:36] <Simon Wilkinson> You probably don't want to merge, ever.
[11:52:41] <Simon Wilkinson> (OpenAFS's workflow is cherry-pick rather than merge based. Our upstream repository has no merge commits in it, and we don't take merges that get pushed to gerrit)
[11:53:41] <Simon Wilkinson> What you want to do is to checkout the local branch, then git fetch (to update your local tree), then git rebase origin/master (which puts all of the upstream changes into the local branch, and moves any local changes that are still relevant after them)
[11:54:32] <kaduk@mit.edu/barnowl> Okay.
[11:54:41] <Simon Wilkinson> Because git SHA1s store not just the change, but also the history, this will change the SHA1 of your patches. But, gerrit has the ChangeId (you are using the change ID hook, I trust) which it uses to keep the association between a commit and an entry in gerrit.
[11:55:54] <Simon Wilkinson> Sorry you've had to fix so much of my breakage ...
[11:57:10] <kaduk@mit.edu/barnowl> Well, it's hard to know that it's breaking when there's no way
to test it ...
[11:57:29] <kaduk@mit.edu/barnowl> ... and I think some of these bugs are much older, anyway.
[12:06:19] <Simon Wilkinson> Oh well, at least I didn't break everything.
[12:06:31] <Simon Wilkinson> We should probably try and get a FreeBSD machine into buildbot when it's done.
[12:07:34] <kaduk@mit.edu/barnowl> That would be nice :)
[12:10:04] <Simon Wilkinson> I just need to get enough spare time to finish the buildbot master, and then I'll start begging for cycles.
[12:57:33] <kaduk@mit.edu/barnowl> If I'm not on a Sun, I don't care about the afs_BozonLock family, right?
[12:57:51] <Simon Wilkinson> Indeed
[12:58:00] <Simon Wilkinson> Sun, and IIRC, some HPUX.
[12:58:39] <Simon Wilkinson> grep for BOZON in src/config for the list.
[12:59:45] <kaduk@mit.edu/barnowl> ... ppc_darwin_80 ?!
[13:00:44] <Simon Wilkinson> ... doesn't have BOZON, from what I can see.
[13:02:04] <kaduk@mit.edu/barnowl> It showed up in my grep.  Anyway, no FBSD show up, which is enough for
now.
[13:20:05] <kaduk@mit.edu/barnowl> I'm having a hard time seeing where the 'p' variable that's passed
as the first argument to afs_setpag() near line 690 of afs_syscall.c
(the DARWIN || XBSD case) is declared.
(My compiler is warning me that I'm passing that argument from an
incompatible pointer type.)
[13:33:29] <Simon Wilkinson> kaduk: 1.5.x or .1.4.x ?
[13:33:36] <kaduk@mit.edu/barnowl> Okay, I guess I see it, now.
1.5.x
[13:34:24] <kaduk@mit.edu/barnowl> Looks like af3_syscall is declared to take an afs_proc_t * in
afs_prototypes.h, but struct thread * in afs_syscall.c itself.
[13:35:01] <kaduk@mit.edu/barnowl> Or ... not.
[13:35:11] <kaduk@mit.edu/barnowl> Not sure what I was looking at, before.
[13:36:11] <kaduk@mit.edu/barnowl> Ah, it was afs_setpag that wanted the afs_proc_t *.
Hm, and an int *retval as opposed to long *.
[13:50:08] --- kaj has become available
[13:50:58] <kaduk@mit.edu/barnowl> ... and that instance of struct thread for fbsd (in afs3_syscall)
goes back to r1.1 in CVS?
[14:15:35] <Simon Wilkinson> We used to have so many compiler errors in that code, that I doubt another one would have been noticed.
[14:18:22] <Simon Wilkinson> http://www.dementia.org/twiki/bin/view/AFSLore/Gsoc2010Ideas
[14:19:08] <Simon Wilkinson> (sorry, wrong window, but feel free to take a look at that page and add your own projects for this year's Soc)
[15:36:45] --- mdionne has become available
[17:37:18] <kaduk@mit.edu/barnowl> I'm not sure I believe the locking in 6c628445.
Also, initializing a mutex whose backing store has not been allocated
doesn't work very well.
[17:38:25] <kaduk@mit.edu/barnowl> It looks like struct mtx is small enough that it could be included
in struct afs_event, which might make that a bit easier.
[17:41:17] --- kaj has left
[18:48:41] --- mdionne has left
[19:42:07] --- Russ has left: Disconnected
[19:47:14] --- Russ has become available
[23:15:29] --- Russ has left: Disconnected