[00:09:27] --- Simon Wilkinson has become available [00:14:09] --- dev-zero@jabber.org has left [00:32:03] --- dev-zero@jabber.org has become available [00:40:39] --- abo has left [00:40:59] --- Simon Wilkinson has left [00:49:37] --- abo has become available [01:56:53] --- dev-zero@jabber.org has left [02:04:27] --- dev-zero@jabber.org has become available [05:12:03] --- mvitale has become available [05:52:00] --- kula has left: Lost connection [05:52:00] --- dev-zero@jabber.org has left: Lost connection [05:52:41] --- kula has become available [05:53:33] --- kula has left: Lost connection [05:57:11] --- dev-zero@jabber.org has become available [06:27:42] --- kula has become available [07:04:48] --- deason has become available [07:33:52] --- Roman Mitz has become available [07:34:48] --- Simon Wilkinson has become available [07:42:39] --- Simon Wilkinson has left [07:42:41] hrm... seems the new mac os x buildbot and the rhel6/64 builtbot are both failing compiles daily [07:51:13] the rhel6 one looks like it's missing a variable expansion on installing rxkad.h or something.... but there's no variable there; that looks odd [07:51:17] I'm probably misreading or missing something [07:54:51] same problem for the macosx build. looks like $? is not being substituted [07:55:17] "/usr/bin/install -c -m 644 /home/buildbot/slave1/rhel6-x86_64-builder/build/include/rx/rxkad.h /usr/bin/install: missing destination file operand after `/home/buildbot/slave1/rhel6-x86_64-builder/build/include/rx/rxkad.h' Try `/usr/bin/install --help' for more information. " sure is a variable substitution error, but i have no clue why [07:56:17] where's the line with the actual variable that should be substituted? [07:57:13] ${TOP_INCDIR}/rx/rxkad.h: rxkad.h ${INSTALL_DATA} $? $@ [07:59:14] (this is the install of the header to the internal include directory, not the dest/install rules) [07:59:34] that from a makefile or from a script? [08:00:15] Makefile. [08:00:28] src/rxkad/Makefile [08:07:08] oh sorry, yeah, reading the wrong install rules [08:10:59] started failing betweeen 69f26ece3c45 and 03b87dffe, I think, for the rhel6 one [08:15:36] so, is it $? or $@ that is failing to substitute? [08:15:55] I'm guessing $? [08:17:22] It's clearly $? [08:20:03] so, either a bug in make or rxkad.h is not "more recent than the target" ? [08:20:41] Yeah, but I fail to see how the latter can be true, since that's the only thing that can trigger the rule at all. [08:21:15] well... [08:21:19] Personally, I'm against gratuitous use of $? [08:21:24] -rw-------. 1 buildbot buildbot 4586 Sep 25 00:08 ./src/rxkad/rxkad.h -rw-r--r--. 1 buildbot buildbot 4586 Sep 6 00:09 ./include/rx/rxkad.h -rw-r--r--. 1 buildbot buildbot 4586 Sep 12 00:09 ./amd64_linux26/dest/include/rx/rxkad.h [08:22:15] Right, so the target exists, but rxkad.h is newer than it [08:22:41] However, this suggests the buildbot is not starting with a pristine tree, which is probably a mistake. [08:22:58] in fact, the "target", if that is ./include/rx/rxkad.h, is dated way way old [08:23:27] It is; the rule's target is ${TOP_INCDIR}/rx/rxkad.h [08:24:25] so, it does a "git fetch", "git reset", then a "git branch" [08:24:33] I thought $? wasn't portable outside of pattern rules, too, but I'm probably mixing that up with something else [08:24:36] then a git rev-parse [08:24:59] then a "git gc" [08:25:02] No, I believe it's $^ that's not portable outside pattern rules. In any case, at least some of the platforms involved clearly use gmake [08:25:33] then git clean -x -f [08:25:51] Right, none of those affect files that git doesn't control. [08:25:58] Oh, but that does [08:26:08] Removing src/rxkad/rxkad.h [08:27:21] Probably git clean should be invoked with -d also, since otherwise I bet it doesn't traverse ${TOP_INCDIR} at all [08:28:03] doesn't look like it is traversing src/include at all [08:29:30] In any case, the way the rule can fire with $? being empty is that src/rxkad/rxkad.h is generated, and running the rule that generates it automatically causes any rules that depend on it to fire. But if there is a "really new" ${TOP_INCDIR}/rx/rxkad.h, then $? will still be empty. Where "really new" here means just slightly newer than "now". ... but that doesn't appear to be the case, from the timestamps you showed [08:29:32] src/include? [08:30:00] No, ${TOP_INCDIR} is include [08:30:35] And yes, since it's an untracked directory, git clean won't traverse it unless you use -d [08:30:39] "Not removing include/" [08:30:39] i know. i was wondering what he was getting at [08:30:58] "Not removing lib/" [08:31:09] which I'm guessing just refers to the directory, not its contents [08:31:15] the other build slaves use dxf, i thought. does your git not have -d? [08:31:24] but regardless, only reference to that in there [08:31:33] the man page shows -d [08:31:39] does the binary? [08:31:43] git clean —help [08:31:47] Remove untracked directories in addition to untracked files. If an untracked directory is managed by a different git repository, it is not removed by default. Use -f option twice if you really want to remove such a directory. [08:31:57] yeah, don't care what the doc says. [08:32:09] git clean --help just shows the manpage [08:32:16] feh. [08:32:33] well, i can tweak the clean step [08:32:34] rhel6factory.addStep(ShellCommand(command=["git","clean","-x","-f"])) [08:32:35] git clean -h [08:32:55] usage: git clean [-d] [-f] [-n] [-q] [-x | -X] [--] ... -q, --quiet be quiet -n, --dry-run dry run -f, --force force -d remove whole directories -x remove ignored files, too -X remove only ignored files [08:33:20] hm. no -e. what did we decide to do to exclude .buildbot-sourcedata? [08:34:53] i'm not going to touch this until other folks are around. the reason this is such a mess is we need to clean but not touch .buildbot-sourcedata. ugh. [08:35:13] later buildslave versions fix it. [08:35:27] hell. you may have a new enough buildslave [08:35:37] think we had determined the buildslave was supposedly new enough [08:35:51] the problem is that -d was removed from the clean rules [08:36:10] -d needs to be put back [08:36:14] well. we can put it back bu ti am not going to do it wholesale. i could do it for those 2 builders. [08:36:24] we can edit further later if needed. [08:37:19] yeah, the mac builder seem sto have the same issue [08:37:21] git clean -h [08:37:51] though, you could just call the correct git on that one [08:38:15] Oh; maybe -d was removed to protect .buildbot-sourcedata It's a bit ugly, but I suggest just moving that into .git [08:38:33] since you're using the system git (1.7.5, I think) instead of the newer version [08:38:40] yes, see, we're trying to avoid custom config [08:38:53] in new buildslaves, the buildslave handles putting it elsewhere [08:39:09] testing new config for macos and rhel6 only. [08:39:32] This is a bit ugly, but... git clean -fx git clean -dfx src rm -rf include lib [08:40:26] well, if both those machines are new buildslaves, i don't care, i can just use -dfx [08:40:53] rhel6 has been installed for a good while, so, it's not "new", depending on your definition of new [08:41:09] where is the "new" git? [08:41:18] on mac os x ? [08:41:24] yes. path. [08:41:43] the bin is /usr/local/git/bin/git [08:41:51] all its supporting stuff is under /usr/local/git [08:42:12] roar:~ cnf$ ls /usr/local/git Library bin contrib lib libexec share [08:42:13] next restart will use that for clean [08:42:35] should I run a buildslave stop / start on that? [08:42:53] the master needs to be restarted, not the slave [08:43:20] i don't want to restart the master again already [08:46:00] speaking of which, do we need solaris 10 sparc and solaris 9 sparc builders? [08:46:15] probably. [08:46:38] well, if we do, I can "probably" supply them... just need to know for sure one way or the other [08:48:21] I may be able to make those appear, too. Or rather, I could probably provide machines, if someone else sets them up remotely. [08:48:38] sets up te buildslave part, that is. anyway, GONE [08:48:57] well, I have some hardware that is either gonna be used for that or is gonna get tossed :) [08:54:18] --- dev-zero@jabber.org has left [09:14:34] --- Derrick Brashear has left [09:30:47] --- dev-zero@jabber.org has become available [10:07:04] --- jaltman/FrogsLeap has left: Disconnected [10:11:38] --- jaltman/FrogsLeap has become available [11:20:39] how did the BOF @ SNIA Storage Developers Conference go ? [11:39:39] --- mvitale has left [11:52:22] --- mvitale has become available [11:57:18] --- natefoo has become available [11:58:18] anyone have any tips for creating new db servers in ec2? it seems to not like the fact that the public IP doesn't exist as a real interface. [12:01:54] have you tried NetInfo with a line beginning with 'f'? [12:01:55] You probably need a NetInfo file explicitly naming the public address. [12:02:05] see http://docs.openafs.org/Reference/5/NetInfo.html near the end of the 'server netinfo' section [12:02:15] ... with a leading 'F', to tell it it's OK to advertise it even though there is no interface with that address [12:02:26] fantastic, thanks. [12:06:05] --- Derrick Brashear has become available [12:08:59] doesn't seem to make a difference. i'm also getting: vlserver: Ubik init failed: problems with host name [12:09:54] what's the host name resolve to? [12:10:39] the external ip. [12:12:11] anything else in the log? [12:12:24] did you add the internal ip, too, or just the external one? [12:12:49] i tried both. [12:12:51] --- ktdreyer has left: Lost connection [12:13:35] here's the entire VLLog when i restart: https://gist.github.com/3783843 [12:15:58] --- ktdreyer has become available [12:18:04] so, there are 3 addresses there; a 54. a 10. and a 128. [12:18:14] what are all of them? [12:18:30] 54 is the ec2 public address, 10 is the ec2 internal address, 128 is the existing db server. [12:20:34] so... the 128. is a separate machine than the 54. ? [12:20:53] oh, yes, sorry for being unclear. [12:21:10] 54 and 10 are the same machine, 128 is a separate machine outside of ec2 that is currently the sole db server for this cell. [12:24:30] so, what is in NetInfo for that log? [12:24:48] 10.211.157.137 [12:24:48] f 54.243.198.183 [12:29:27] the log messages are the same regardless of whether i have NetInfo, including that top "does not exist" for the 54 address, which is a bit odd. [12:30:12] like it's not reading netinfo, but it's in the right place - i have fhs paths and it's in /etc/openafs/server [12:31:50] it goes in /var/lib/openafs/local these days, I thought [12:31:57] or /etc/openafs/server-local a long long time ago [12:32:19] whatever the /usr/afs/local equivalent is; I thought /etc/openafs/server was the /usr/afs/etc equivalent [12:32:42] hah, so it is. [12:32:49] works fine when i put it there. ;) [12:33:25] thanks! [12:35:22] cool [12:35:46] you weren't reading a document that said to put them in etc/openafs/server, were you? [12:36:36] if that was just a mistake, then good (people do that all the time); just wanna make sure there's not a howto or something saying something wrong [12:37:58] no, just the page you linked, which uses transarc paths. [12:38:06] so i was assuming. [12:46:01] It's a little wonky. There isn't a direct mapping from transarc paths to fhs paths, because internally there are actually more directories, some of which default to the same thing. But which ones are the same depends on whether you are using transarc or fhs [12:51:12] The NetInfo file goes in afslocaldir, which defaults to /usr/afs/local or ${localstatedir}/openafs But not everything from /usr/afs/local goes to ${localstatedir}/openafs. The BosConfig file lives in afsbosconfigdir, which also defaults to /usr/afs/local in transarc mode, but ${sysconfdir}/openafs in fhs mode. [13:02:26] heh. [13:02:27] makes sense. [13:07:33] we could still have a "documentation" page or something showing that mapping; I don't think we have anything like that [13:08:08] I can imagine it may be difficult to keep up with where all various packages put things... but I'd hope it wouldn't change too much [13:09:30] damnit. "Error 405" ate my message [13:09:49] anyway, the right answer is a krb5-config style executable which emits any dirpath based on your actual install [13:12:33] that would be nice; sounds easy to do [13:13:02] that alone doesn't tell people e.g. "where does NetInfo go" unless the documentation says that netinfo goes in afslocaldir [13:14:08] if you also print out what the transarc path would've been, that maybe provides at least a hint when you're looking at documentation that just says /usr/afs/local [13:14:38] afs-dirpath —server NetInfo [13:14:48] /usr/afs/local/NetInfo [13:16:12] okay sure; thought you meant just the dir [13:16:14] nice [13:16:35] I was thinking each file would be harder to maintain, but I guess they don't change much and there aren't _that_ many... [13:21:49] in the ideal case you don't need to maintain them. they're enumerated in dirpath already. just walk the array until you find it. the only hard bit is coding what's client versus server versus universal [13:22:00] and that can be done by knowing what prefix is what [13:22:20] afs-dirpath doesn't need to be fast, so walking the array? who cares. [13:52:07] --- abo has left [13:58:19] --- abo has become available [14:05:27] --- jaltman/FrogsLeap has left: Disconnected [14:08:20] --- jaltman/FrogsLeap has become available [14:16:41] --- Derrick Brashear has left [14:58:09] --- Derrick Brashear has become available [15:01:22] --- mdionne has become available [15:35:18] --- mvitale has left [15:49:43] --- deason has left [17:21:00] --- ktdreyer has left [17:23:00] --- ktdreyer has become available [17:35:56] --- ktdreyer has left [17:38:55] --- ktdreyer has become available [18:48:19] --- mdionne has left [23:02:59] --- dev-zero@jabber.org has left [23:12:18] --- jaltman/FrogsLeap has left: Disconnected [23:12:31] --- jaltman/FrogsLeap has become available [23:14:36] --- dev-zero@jabber.org has become available [23:30:56] --- dev-zero@jabber.org has left