[01:35:15] --- jaltman/FrogsLeap has left: Disconnected [02:00:06] --- jblaine has left [02:06:42] --- jblaine has become available [05:24:33] --- summatusmentis has become available [05:25:25] is there a preferred location for new rx test functionality? [05:55:36] --- jaltman/FrogsLeap has become available [08:10:38] --- deason/gmail has become available [08:22:38] deason: I apologize for being snarky re: cvs.html last week. I was frustrated. [08:25:11] are there mechanisms in place for the TAP testing stuff to mock a required setup? [08:30:51] jblaine: np, sorry if I seemed overly harsh [08:31:19] summatusmentis: like to create a miniature cell? simon said that he had something in the rxgk repo [08:32:33] deason: right, I remember simon saying that. I'm not sure I need the full functionality, but I'm trying to write a test for some rx functionality [08:32:42] (see change 4714 if you care) [08:33:31] and so, either I need to mock some rx stat data to test against, or the test needs to be able to access current running client data [08:35:46] and it's all different enough from the JUnit java testing we did in school that I'm not 100% how to go about it [09:04:25] well, you'd need to initialize rx but you don't necessarily need to actually do anything with it [09:04:53] you could just make some functions to add a peer and manually set the stats data, and then the test is calling that function to read it back and see if it matches [09:05:01] mmm, nothing quite like debugging someone's tcsh script [09:05:40] (er, by "do anything with it" I mean, actually make any rpc calls or hit the net, etc) [09:06:15] deason: by add a peer, you mean construct an rx_queue and essentially mock out the whole structure? [09:07:22] no, I mean add a function to librx that says "add this peer" [09:07:24] or no, no [09:07:54] you don't need to do that; if you rx_NewConnection() to a peer, the peer will be created, but the information provided doesn't need to be valid, really [09:09:04] then make an rx_SetPeerStats() or something; I'm not familiar with the data involved, so I don't know how you'd be setting it [09:10:13] or you can mock the whole structure; you may know better than I how difficult that is and so whether that's a better way [09:10:49] hrm... alright, I think [09:29:28] --- todd has become available [09:30:00] TK.media 536871366 RW 96182991 K On-line [09:30:10] maybe I should think about creating some separate volumes for this data. heh. [09:34:35] only 96GB ? [09:34:37] tiny. [09:39:25] --- phalenor has left [09:39:31] --- phalenor has become available [09:48:49] --- rra has become available [09:59:21] hmm, so, while trying to understand how libafs paging works, I'm not really understanding our current behavior on (I think) all unix platforms.... if someone could tell me how much of this is intentional and/or unavoidable, it'd help: [10:00:08] on all platforms, we don't invalidate pages for a file on a callback break, so any mapped pages can see stale data if another client writes to the same file [10:00:38] but we do invalidate the pages on, say, a read, so mapped pages do see changes if someone else on the same client starts a read (or presumably, creates a new mmap) [10:01:24] it also doesn't look like we invalidate pages on dcache entry eviction anywhere except solaris, which suggests to me that stale data could occur that way, too, though it's harder to reproduce that so I'm not sure [10:05:14] I'm not sure why we can't osi_FlushPages or something in ClearCallback, and do something similar to solaris' osi_VM_GetDownD in afs_GetDownD everywhere [10:05:59] the comments for osi_VM_GetDownD suggest it's not necessary for non-solaris, but I have no idea why because I have no idea wtf this comment is talking about: * This function only exists for Solaris. For other platforms, it's OK to * recycle a dcache entry without invalidating pages, because the strategy * function can call afs_GetDCache(). [11:03:46] --- jaltman/FrogsLeap has left: Replaced by new connection [11:03:47] --- jaltman/FrogsLeap has become available [11:04:14] --- jaltman/FrogsLeap has left: Disconnected [11:05:22] --- jaltman/FrogsLeap has become available [11:15:18] buildbot is back [11:15:32] Woo! [11:33:01] How much backlog does it have to chew through? [12:03:24] --- deason/gmail has left [12:26:40] --- deason/gmail has become available [12:57:56] --- jaltman/FrogsLeap has left: Disconnected [12:58:08] --- jaltman/FrogsLeap has become available [13:24:23] --- Jeffrey Altman has become available [13:29:26] --- Jeffrey Altman has left [14:18:20] --- jakllsch has left [15:22:22] --- deason/gmail has left [15:56:04] --- kula has left [15:58:56] It looks like I will need to manually take action to get buildbot to build a change I submitted while it was off the net? What's the best way to do so? [16:22:54] --- rra has left: Disconnected [16:39:50] --- Russ has become available [16:39:58] Anyone around who can run the openafs-devel moderation queue? [16:41:05] --- jaltman/FrogsLeap has left: Disconnected [16:46:43] --- kula has become available [17:16:34] Are there any clients where we currently give each volume a different device ID? [17:52:19] --- shadow@gmail.com/barnowlA109197F has left [17:53:50] --- shadow@gmail.com/barnowlA109197F has become available [18:09:45] --- steven.jenkins has left [18:11:24] --- steven.jenkins has become available [18:45:39] --- steven.jenkins has left [18:45:40] --- shadow@gmail.com/barnowlA109197F has left [18:46:37] --- steven.jenkins has become available [18:47:08] --- shadow@gmail.com/barnowlA109197F has become available [19:05:13] --- steven.jenkins has left [19:05:33] --- steven.jenkins has become available [19:20:19] --- jaltman has become available [19:21:46] to force a build, you can rebase and push the change back to gerrit [19:21:54] I can't moderate [19:22:38] the windows client treats every volume as a distinct device. mount points are reported a reparse points to the IFS [19:24:07] Hmm, buildbot seems to have picked it up on its own, actually. And it compiles! [19:24:44] it periodically looks for things that were missed. or the mail notification for the previous submissions finally arrived [19:24:50] Is there consensus that reporting each volume as a separate device would be a good thing, if we have the capability to do so? [19:24:59] I think so [19:25:28] Nifty. (AFS came up on a thread on freebsd-fs regarding 64-bit inode numbers.) [19:31:24] --- jaltman has left: Disconnected [19:31:50] --- jaltman has become available [20:03:28] --- Jeffrey Altman has become available [20:15:40] --- jaltman/FrogsLeap has become available [20:17:19] --- Jeffrey Altman has left: Replaced by new connection [20:17:20] --- Jeffrey Altman has become available [20:17:47] --- jaltman/FrogsLeap has left: Replaced by new connection [20:17:48] --- jaltman/FrogsLeap has become available [20:59:28] (I feel kind of weird just submitting a build system change, albeit minor, to gerrit without wider exposure for commentary.) [21:11:27] submit the change to gerrit and post a description and request for comments to openafs-devel [22:36:11] --- jaltman has left: Replaced by new connection [22:36:12] --- jaltman has become available [23:57:35] --- Russ has left: Disconnected