[00:42:28] --- dev-zero@jabber.org has left [00:44:03] --- reuteras has left [00:44:41] --- reuteras has become available [01:09:14] --- abo has left [01:09:33] --- abo has become available [01:41:21] --- Jeffrey Altman has left: Replaced by new connection [01:43:35] --- dev-zero@jabber.org has become available [02:29:40] --- Russ has left: Disconnected [02:52:53] --- Claudio Bisegni has become available [02:54:00] --- Claudio Bisegni has left [03:01:50] --- cclausen has become available [04:33:28] --- haba has left [04:52:54] --- cclausen has left [05:42:41] --- Jeffrey Altman has become available [06:18:16] --- cclausen has become available [06:26:14] Yeah, that's because the poster was not subscribed, and it was in the moderation queue for that long. Convince the world to send us less spam, and moderation will be less painful and thus happen more frequently. [07:05:57] --- deason has become available [07:16:29] --- reuteras has left [08:09:25] --- haba has become available [08:12:22] --- matt has become available [08:13:17] --- sxw has become available [08:16:26] --- matt has left [08:28:09] --- matt has become available [09:19:21] Looks like afs_fetchstore.c doesn't build in UKERNEL? atm [09:20:01] I get some unused vars warings from it in the KERNEL build, also [09:20:39] All of the patches have been verified somewhere, so I suspect it not building may be platform dependent. [09:20:51] linux [09:20:56] Doesn't surprise me if there are unused vars warnings - we're not verifying with warning checking turned on. [09:21:08] It's possible that Derrick has pushed patches based on testing on Mac OS X only. [09:21:54] Once I get some spare cycles, buildbot should help fix this, at least for platorms where we can build quickly. [09:21:56] I am building with some header changes, but shouldn't be cause of this, but I'll drop back and try again, just to check. [09:22:35] I haven't tried to build on Linux for a while ... [09:26:29] russ: indeed, my use of typedef was a red herring. all that's required moving the struct keyword into the rhs, the rest is as I stated [09:27:11] I do have a preference, generally for using typedef, rather than #define for type assignment. [09:29:06] I doubt that's ruled out, I'm just trying to correct one problem at a time, and #define is what's there now [09:29:17] sure. [09:30:43] there's a bit of error pasting conditional code in fetchstore.c [09:33:49] is someone around to look at afs_fetchstore.c:627? shouldnt newcall label be outside the afs_Analyze loop? [09:34:40] I mean, nocall [09:35:14] --- haba has left [10:13:32] Actual stop in the build is just that the target of the goto is the end of the do block. That seems to mean that AFS_64BIT_CLIENT is !defined for UKERNEL, but is for kernel build. I wouldn't have expected that. [10:22:50] --- matt has left [10:34:24] --- haba has become available [10:52:41] --- Russ has become available [10:53:51] matt: We're halfway through applying a set of patches, so it should get better as soon as work stops being an endless nightmare and I can get through the rest of them. [10:53:53] There are known unused variable warnings that will go away with Felix's latest patches as well. [10:54:09] Sorry that the process of applying that whole stack has been so slow. [10:54:43] All the patches I applied did build on Linux, but Derrick may have pushed more of the set. [10:55:41] I think Felix's patch set has taught us a lot about the issues with gerrit and long chains of dependent patches. [10:57:20] it taught us they suck [10:57:48] Indeed. [10:57:59] --- matt has become available [10:58:02] They suck for Felix, more than they suck for us. But they truly suck. [10:58:28] The question is, what should we do instead? [10:58:42] Well, besides "don't develop a ton of code out of the tree," which is the long-term solution. [10:58:43] I think we encourage people to push patch sets into gerrit in smaller chunks. [10:58:43] cry [10:59:13] I don't think making change harder is the answer. [10:59:24] I suppose some more aggressive squashing could have reduced Felix's patch set by about half, although it was in nicely digestible chunks for someone like myself who isn't that familiar with the code. [10:59:41] He did a great job putting moving code and cleaning up code into different patches, which is important. [10:59:56] I really liked the granularity of Felix's patch set. I wish every patch set we received was split up in that way. [11:00:00] I wasn't complaining. [11:00:26] One of the problems is that we're doing design review at the same time as code review. [11:00:28] The thing is, that whole set actually doesn't change a lot in terms of getting us to RxOSD, so if he pushed it even slower, wouldn't that slow things down even more? [11:00:31] That is true. [11:00:37] some of the granularity was bad. most wasn't [11:00:59] If we'd hammered out how the method redesign would work first, the code review probably would have been slightly faster. [11:01:10] there were a couple of cases where it looked like change A, change B, the rest of the stuff that belong in change A, for instance, for a set of 3 commits [11:01:16] Yeh. [11:01:22] Yes, there were a few of those. [11:01:27] And there were dependent changes that actually weren't. [11:01:45] pushing it slower would only have benefit in not forcing him to rebase changes and repush to gerrit. he could rebase before the first push. otherwise, no value [11:01:45] --- dev-zero@jabber.org has left [11:02:01] That has value, though. [11:02:20] In practice, his rebasings have mostly not been enough. [11:02:20] The changes in 2.0.19 will help hugely, because you will no longer have to target changes at specific change IDs. [11:02:26] I've been rebasing each patch as I submitted it. [11:02:34] --- abo has left [11:03:06] rebasing each as you submit means no review is possible until we accept the previous and you rebase, which does lose time [11:03:15] --- deason has left [11:03:18] --- abo has become available [11:03:20] --- stevenjenkins has left [11:03:20] If you have a tree of changes, and you change the first one, and all the rest are rebased, you will just be able to do push to refs/for/master, and not need to use refs/changes at all. [11:03:29] if 391 and 392 could get some comment i'd appreciate it. notably 392, but... [11:03:32] --- deason has become available [11:03:38] Basically, what I've had to do with nearly all of them is: git fetch ... && git rebase --onto master FETCH_HEAD && git checkout tmp-verify && git merge [11:03:41] Because we cherry pick, there will always be rebasing issues [11:04:18] Because if I try to git pull the change, the change ends up as nothing but merge conflicts. [11:05:18] --- deason has left [11:05:25] --- deason has become available [11:06:47] --- stevenjenkins has become available [11:19:51] --- Rrrrred has left [11:28:32] --- sxw has left [11:43:21] --- sxw has become available [11:45:09] --- sxw has left [12:05:47] --- dev-zero@jabber.org has become available [12:31:12] --- dev-zero@jabber.org has left: offline [13:57:01] --- haba has left [14:57:07] --- Rrrrred has become available [14:57:55] --- asedeno has left [14:57:55] --- asedeno has become available [14:58:00] --- asedeno has left [14:58:15] --- asedeno has become available [14:59:37] --- haba has become available [15:01:27] --- matt has left [15:16:56] --- deason has left [15:39:04] --- andersk@mit.edu/dr-wily has left [16:07:08] --- cclausen has left [16:39:25] --- deason has become available [17:01:53] --- asedeno has left [17:01:53] --- asedeno has become available [17:01:57] --- asedeno has left [17:02:10] --- asedeno has become available [17:10:07] --- Russ has left: Disconnected [17:10:14] --- Russ has become available [18:38:36] --- dev-zero@jabber.org has become available [18:53:09] --- Russ has left: Disconnected [19:11:18] --- Russ has become available [20:46:44] --- Jeffrey Altman has left [21:16:13] --- Jeffrey Altman has become available [22:01:57] --- deason has left [23:16:59] --- Russ has left: Disconnected