=== Makyo is now known as Makyo|out [13:22] gary_poster: I would like to verify that the loading message branch is doing the right things. We can have a call to discuss if you want. [13:23] benji, I was trying to reply. I had a security breach in my accounts yesterday, which made me actually start to use LastPass, which is slowing me down a bit as I try to understand it. All I wrote was this: [13:24] app/index.html:62: Trying to connect to the Juju environment. [13:24] Initially this is about loading JS. It's only about connecting once the app has [13:24] started. The login shouldn't show until we've connected. [13:24] Does that make sense? Happy to call. [13:25] gary_poster: I think I took that wording from the UI mockup. I'm happy to change it. How about "Loading the Juju GUI."? [13:26] +1 benji. We can then use moxckup's wording once we are connecting [13:26] gary_poster: I don't undestand that bit. Should there be another box that pops up with a connecting message? [13:27] benji, I suggest changing text [13:27] gary_poster: -1 on breach. +1 on lastpass though! [13:27] :-) yeah [13:27] yeah, sorry to hear that; I hope it has been easy to clean up [13:28] so far not bad. Takes time to get everything back using the new passwords, and I have so many google passwords, and so many places they are remembered that it is the most annoying bit [13:29] gary_poster: by the way, I found a spinner generator and made a spinner that looks like the one in the mockup... , [13:29] benji, heh, was that the one in the branch and I should have actually looked at it? was about to do that when I went down a LastPass rabbit hole [13:29] will look [13:30] yep, it's in there [13:30] I need coffee. I'll be back in a sec. [13:38] benji, spinner is not like js. spinner js is also 1/6 the K weight if gif [13:38] of [13:39] benji, also, in non-chrome browsers, should only show spinner message once "your browser is not supported" message has been clicked [13:43] gary_poster: i could not reproduce that routing bug so i'm glad to see your comment on it. i've marked it invalid. [13:43] gary_poster: good info, thanks [13:43] thanks bac, benji [13:47] gary_poster: preferences on next card i should grab? [13:48] bac, if you don't have a Go Juju dev environment, that would be good to make. Otherwise I like bug 1117888 [13:49] or bug 1119412 [13:49] good morning [13:49] morning hatch [13:49] did you start that styling bug you mentioned or did it end up being a rabbit hole? [13:50] yo _mup_. You going to translate those bug numbers? [13:50] I 'started' on it but didn't get far, once the coffee hits I'll be resuming [13:50] :-) ok hatch. If you want to pair I can help or someone else can [13:50] did I put the task on the kanban in the right place? [13:51] nah, it is a slack task, and you are working on it, so it goes in Slack -> Coding -> Active, hatch. I moved it [13:51] got it [13:52] bazaar kind of feels like svn gone distributed with the way it handles branches [13:53] hatch yeah. there are a bunch of ways to use it though. This is the easiest entrance, I think. There are other approaches that are more like git, though they are a bit clunky IMO in the bzr world [13:53] We can explore at a sprint [13:54] BTW, speaking of sprint, everyone, hazmat sez that location is still not set, so that's why we have not had a grand ticket buying exercise [13:55] jujugui ^ [13:55] ooh, I could use that for my daily call pings! [13:55] gary_poster, location close to set (atlanta), but haven't gotten final confirmation yet. trying to figure out accomodations. [13:55] cool hazmat thanks [13:56] hmm I'll have to figure out what to do for cell data while traveling [13:56] I'd need a second job to pay for it if I didn't find a plan or buy prepaid heh [13:56] hatch, prepaid sim card from tmobile, att should do it [13:56] $5/MB [13:57] assuming unlocked phone [13:57] yeah it's not....but I'm debating picking up a nexus 4, (or unlocking this one) [13:57] hatch, there are some co-located branch add ons (ie multiple branches in the same dir) .. cobzr is pretty popular in the juju-core aka backend team. [13:58] gary_poster: isn't it "guihelp"? [13:58] benji: yeah. i ding on both. [13:58] benji, we should have both.. one is meant for #juju [13:59] benji, yeah, I could do both. jujugui is our ping for customer service help in #juju. guihelp is our ping for getting attention locally. By using jujugui I am making sure that we have the customer service one connected for everyone :-) . I also realized a while ago that we could only have jujugui, so in my mind that's how I think of it [13:59] I hilight on one and make a sound on the other. Making a sound on both seems a bit interruptful for a 10-person team [13:59] juju+gui there gui+help here.. [14:00] yeah.. collapse to one would work fine [14:00] +1 on fewer if we are going to treat them as synonyms anyway [14:01] cool, I'll mention it on call [14:01] do we track time on tickets? [14:01] no hatch [14:02] * benji practices Movie Voice: "Coming to a theater this summer... a space thriller like nothing you've seen before: No Hatch" [14:02] :-) [14:04] clearly that would be an awesome movie [14:06] :) [14:08] so waht is the proper procedure for working with the less file? Do I run make after every change or can I just run less manually? [14:08] I see the make includes recess [14:09] hatch, make devel will automatically notice changes and regenerate [14:10] so edit and reload === teknico_ is now known as teknico [14:21] gary_poster: it seems we have a regression, both in trunk and uistage: "Uncaught TypeError: Cannot call method 'get' of null" while trying to destroy a service (from the svc menu or from the svc detail) [14:22] frankban, :-/ ok. bac, ^^ do you have bandwidth? [14:22] thanks frankban [14:32] gary_poster: yes. i don't understand the context of your question though. [14:33] gary_poster: oh, you mean "brad, look at this regression please." [14:33] bac, yes :-) [14:33] gotcha [14:33] * bac looks [14:33] thanks [14:54] need.....more.....ram!!!! [14:54] I remember upgrading to 4MB [14:54] now 8GB isn't enough :) [14:58] frankban, gary_poster: that regression has been around for 10 days or so. :( [14:58] bac, :-/ [14:59] bac: :-( [14:59] bac, easy to fix? [14:59] easy to test? :-) [14:59] gary_poster: dunno, just traced it back to the broken revision so now i'm trying to see what it did [14:59] frankban, gary_poster bac - Current landing branch may fix. === Makyo|out is now known as Makyo [14:59] I was just going to say...when you guys find things like this do you write a test for it? :) [14:59] Whoops. [14:59] ack cool Makyo [15:00] yes hatch :-) [15:00] Makyo: so a deleted service is still in the set of services but has no model. does that sound like something you just fixed? [15:00] bac, yeah. [15:00] cool [15:00] great [15:00] bac, but by accident. Will do a quick test branch after lbox finishes. [15:01] Makyo: do you want to write a test for it or would you like me to? [15:02] bac, if you'd be willing, that'd be excellent :) [15:02] hatch, how many VMs do you have in there? :-) [15:02] Makyo: sure [15:02] teknico: just Ubuntu and Win 8 but I can't boot the win 8 one up because OSX Is using too much ram [15:04] "Used 7.98GB" - since I have paging turned off I better not use those extra 20MB or something is going to crash ;) [15:04] heh [15:05] gary_poster, do you have 5 mins for me to "quickly" present the landscape integration designs I've just emailed you? [15:05] bac, branch merged. [15:05] Makyo: thx [15:05] goodspud, sure! was about to approve via email, with a mention that we will need "Unit Landscape" too [15:05] going to jujugui [15:14] goodspud, the attachments didn't make it.. can you send a link [15:15] hazmat, give me five and I'll upload them to Drive... [15:28] jujugui call in 2 [15:29] hazmat, emailed you a link. The images have some further exploration of the interface but focus on the Landscape integration aspects [15:29] goodspud, thanks [15:29] ^^ that's the new version of bac bcsaller_ benji frankban goodspud hatch hazmat Makyo teknico announcement :-) [15:29] so please set up the ping if you have not already [15:30] oh it's dinging [15:31] hatch bac Makyo starting without you, come on by [15:50] Makyo: data point: I don't see the weirdness on trunk that you guys see. [15:51] benji, thanks. How is your workspace set up? I have mine set up like http://doc.bazaar.canonical.com/beta/en/user-guide/organizing_your_workspace.html#feature-branches [15:53] Makyo: yep, mine is set up like that [15:56] Hmm, alright. Just wondering given that our makefile touches bzr [15:59] hazmat, what juju IRC channel did you mention again (for asking about MongoDb), and on what server? [16:01] I guess it's #juju here on freenode, right? [16:02] teknico, #juju-dev [16:02] I thought you mentioned that one, but could not find it [16:03] ok, I'll try joining anyway :-) [16:06] gary_poster, auth stuff was merged fwiw [16:06] cool hazmat [16:21] bcsaller_, a new checkout of trunk works with make devel. Tried it in a separate folder, though. Will try replacing my existing one and see if that helps. Not sure what would've gotten stale, though. [16:21] Makyo: me either [16:25] bcsaller_, also works just deleting and rebranching trunk in existing repo tree. [16:25] ^^^ guihelp - anyone else that had that issue with running trunk [16:25] Makyo, does make clean && make devel work in old trunk? [16:25] hazmat, no :/ [16:26] hazmat: that didn't work for me [16:26] I'd like to understand the cause here [16:26] meld the two directories [16:27] re meld fwiw - if there isn't a sig delta there [16:27] whoops [16:27] -> http://meld.sf.net [16:28] bcsaller_, Makyo what's the symptom? [16:28] back button broken? [16:28] hazmat: no, Makyo posted the link before, an error and broken rendering [16:28] the site just won't function [16:29] make devel on trunk seems to be working fine here [16:29] * hazmat joins the line stoppage [16:30] anyone know this error.. http://paste.ubuntu.com/1639863/ looks like a nodejs inotify issue around the css watch? [16:31] hazmat: yes, I started seeing that as well [16:31] that feels like a newer dep issue [16:32] * hazmat does a make clean-all [16:32] I *think* its not releasing the watch handle properly, but it could be that we just crossed some ulimit number on recent builds [16:32] no reason we should be hitting that [16:32] we don't have that many files to compile.. [16:33] hazmat, hatch points out for landscape work that session storage will be per tab, while session cookie would work across tabs. This won't make a huge difference but could be nice for some stories, like gui -> landscape -> new tab for gui. Only downside is we have to think a bit harder about security. https should make most common concerns fine though. I don't see any issues with shared domain bits. thoughts? [16:34] gary_poster, cookie from who? [16:34] hazmat stash cookie with auth bits [16:35] from ourselves [16:35] gary_poster, is there an api that let's the browser set its own cookies? [16:35] y hazmat [16:35] oh. setCooke [16:36] also hatch tells me sessionStorage in spec is per top-level domain [16:36] cookie is per full domain [16:36] http://dev.w3.org/html5/webstorage/#the-sessionstorage-attribute [16:36] full domain works better for us (e.g. aws) [16:36] ""Each top-level browsing context has a unique set of session storage areas, one for each origin."""" [16:37] YUI has a cookie module as well if we choose to go this route http://yuilibrary.com/yui/docs/cookie/ [16:38] top level browsing context doesn't reference domains afaics [16:38] bcsaller_, make debug and make prod seem to be working in oldtrunk, so perhaps something with a stale file as served by node [16:41] gary_poster, hatch yeah.. the separate per document nature of session storage defeats the purpose of it. cookies or localStorage both work.. re cookies, the only concern is non-https gui usage.. [16:41] but we're really trying to stop that usage for any real world usage in preference of minimally a self-signed cert [16:41] agreed hazmat. [16:42] local storage means client password potentially on disk [16:42] gary_poster, so if i go to a new tab against the same domain do i get back my own cookie? [16:43] y hazmat [16:46] * hazmat explores a bit more [16:51] gary_poster, only concern is that it seems quite strange to send the cookie out with plaintext creds, at min we'd have to require ssl unless we're in.. if we have things like js error collection or analytics, it could easily accidentally sniff those, ditto is possible for logs on the static resource server [16:52] gary_poster, the per domain nature of storage isn't nesc an issue, we can namespace creds using env uuid [16:52] hazmat, per domain: we may have misread. agreed either way not an issue [16:52] gary_poster, i think the per domain is accurate wrt to localstorage [16:52] if js runs on page, both approaches are flawed [16:53] I mean, foreign js [16:53] yeah any js from the same domain will have access to both [16:53] gary_poster, not nesc.. but std browser env stuff can get pushed out.. [16:54] didn't follow last sentence hazmat sorry [16:54] gary_poster, hatch re third party js, the common case is either analytics or error tracking js, typically from different domains [16:54] hazmat: so your worried that the user will access the page via http and the data will be sent clear over the wire? [16:54] that's valid [16:55] sessionstorage won't have that issue because we explicitly need to send it [16:56] gary_poster, things like cookies are common access for analytics, or error collection, storage is not and easily audited [16:56] ack hazmat [16:56] agreed [16:57] hatch, we're trying to minimize http, i hope we can get farther along that road. but say we start hosting to the gui on cdn for better speed access, now we're at the mercy of a third party to complete the loop on security [16:57] yep good point [16:57] so then we should pop landscape open in a new window [16:57] well, we will send the password in the clear over the websocket in that case [16:58] we need https/wss [16:58] gary_poster, cdn is independent of http/https.. we own the websocket code.. and we can promote https there [16:58] gary_poster, even if the cdn is https, we'd still be dependent on the third party to close the security loop, because they could simply log the cookies, even by accident [16:59] true [16:59] hazmat, convinced by sessionstorage. hatch? [16:59] yep sessionstorage looks like the best approach [16:59] cool, thanks both [17:01] i was wondering if we could get away with localStorage to avoid the tab issue.. but realistically we don't want the user to have multiple gui tabs open.. its a sig. waste.. [17:01] gary_poster: remember that permission denied issue I had yesterday - I think it's because npm was installed via sudo :/ [17:02] agreed hazmat. so you agree tab/window for opening launchpad? we can use same name so we don't open billions on repeated use [17:02] s/launchpad/landscape/ [17:02] hatch ^^ :-P [17:02] haha [17:03] hatch, npm is supposed to be installed in system [17:03] gary_poster, we'd always get two.. we can't control the name on the first one [17:03] via apt [17:03] true :-( [17:03] hatch, via chris-lea node js ppa [17:04] (as in hacking doc, as you were looking at) [17:04] yeah I blindly copied that line [17:04] which installs npm under sudo [17:04] so every time I run make devel [17:04] that line is fine [17:04] I need sudo [17:04] :-/ [17:04] not me. package just changed... [17:05] so I need to 'un sudo' npm :) [17:05] hatch, is this same for you? [17:05] $ ls -l /usr/bin/npm [17:05] lrwxrwxrwx 1 root root 21 Nov 5 14:52 /usr/bin/npm -> /etc/alternatives/npm [17:05] and [17:06] $ which npm [17:06] /usr/bin/npm [17:06] yep [17:06] uh. [17:06] I'm logged in as my user though [17:06] yeah, supposed to be [17:06] I can run npm without sudo [17:07] yeah, good, that's what I'd expect [17:07] but running make devel throws an npm error about crypto something [17:07] hatch, can you pastebin the error [17:07] hatch, pastebin? http://pastebin.ubuntu.com/ commonly used fwiw but whatever works [17:07] hatch, incidentally there's a nifty cli for pastebinit.. apt-get install pastebinit [17:08] you can pipe a file into or pass a file as arg [17:08] oh that's cool, does it give you an option to delete these after? :) [17:08] who wants to delet anything? In perpetuity, and beyond! [17:10] http://pastebin.ubuntu.com/1639976/ [17:11] hatch ls -l output at top of branch? [17:12] http://pastebin.ubuntu.com/1639979/ [17:13] I'll create a new branch and run make devel again to get the logs [17:14] hatch, you probably sudo make at some point.. [17:14] it seems like your perms aren't quite right [17:14] hatch, not clear. you already tried make clean, I think, yesterday? would be interesting to find what is owned by root or whomever. You could try sudo chown -R pihach:jpihach . [17:14] i'd suspect some root owned files.. you could also adjust via chown -R : checkout_root_dir [17:15] :-) [17:15] https://gist.github.com/hatched/171fd627c3f757d4111d [17:15] * hazmat wanders back to the line stoppage [17:16] that's thrown right at `npm install` [17:16] hatch, yeah.. definitely ran npm as root accidentaly.. [17:16] hatch, chown -R jpihach: ~/.npm [17:16] and then run again [17:16] you don't need to prefix make with sudo [17:16] it will ask for sudo if it needs it [17:17] ahh there we go - on osx I had to install npm without sudo for it to work properly [17:17] thanks [17:18] hatch, you are using makefile on OS X directly, not in vm? [17:18] no no I'm in VM now, I was referencing what I previously had to do on osx [17:18] oic [17:21] fwiw, make devel is working fine for me [17:21] looks like I need 'shelltoolbox' for lbox propose -cr [17:21] going to have to add that to the HACKING doc [17:21] ImportError: No module named shelltoolbox [17:22] thanks hatch. python-shelltoolbox. Actually, one sec [17:22] bcsaller_, the emfiles on watch thing is concerning [17:23] Makyo, bcsaller_, basically trunk make-devel is busted then afaics [17:23] wfm :-/ [17:23] gary_poster, make clean-all && make devel works? [17:23] hazmat: yes, I can try to track it down later [17:23] hazmat was fresh checkout. can try that too [17:24] hazmat, bcsaller_, it works for me with a fresh checkout, yeah. Exploring more. [17:24] hatch, yes, go for python-checkout [17:24] so the other local state we have is in ~/.npm [17:24] ah [17:24] python-shelltoolbox hatch [17:24] both? ok [17:24] line 53 of HACKING hatch. no only python-shelltoolbox [17:24] hatch, later [17:24] sorry [17:27] ok gota find PPA for that [17:28] before I install.... https://code.launchpad.net/~yellow/+archive/ppa/ [17:28] that's the proper python-shelltoolbox ? [17:28] hatch, no, use juju ppa [17:28] hatch sudo add-apt-repository ppa:juju/pkgs [17:28] that belongs in the hacking doc as well [17:28] you should use that version of juju anyway [17:28] agreed [17:29] hatch ^^ [17:30] hazmat, bcsaller_ , Makyo, fwiw, all of following worked: fresh checkout, make devel, make clean-all, make devel . Agreed that clean-all apparently misses some local state that is corrupt. My trunk has been around for awhile and is ok [17:30] I mean I have a separate, older one [17:30] I think [17:30] doublechecking [17:30] * hazmat tries fresh checkout [17:30] my trunk is definitely feeling a bit busted [17:31] i still suspect we're getting different deps and possible caching in ~/.npm due to inexact version specs [17:31] would prefer exact version specs msyelf [17:32] still working to verify though [17:32] gary_poster: we might have found it, I think its app/templates.js [17:33] some old version was still being used and wasn't sync'd with the current code, it should have been regenerated on server startup though and wasn't, even after a make clean [17:33] gary_poster, we don't control transitive deps though [17:33] although i would hope that's not the issue [17:33] bcsaller_, did that also resolve the em watch error? [17:33] ugh, true hazmat. we could if we specified all of them [17:33] yuck [17:34] hazmat: I don't get that everytime, only if the systems been running for a while [17:34] bcsaller_, aha, thanks [17:34] hazmat: which is why I think there is a descriptor leak [17:34] * hazmat stops barking at the wrong tree [17:35] wow propose takes a long time [17:35] bcsaller_, this is my first gui dev in a bit, so it could be a leak unrelated [17:35] need more powa!! [17:35] hatch, i always run lbox commands with -v [17:35] oh that's a good idea [17:36] Yeah, tracked the error from console out to stale templates. [17:36] Should clean get generated files like templates.js? [17:36] Cool bcsaller_, Makyo thanks [17:36] Makyo, definitely [17:36] I thought it cleared out all build-* [17:36] which ought to do the trick? [17:36] templates.js is stored in app/ [17:37] Makyo, that's old [17:37] should be anyway [17:37] I think... [17:37] checking [17:37] yes Makyo that's old [17:37] app/templates.js should not exist [17:37] nothing builds it any more there AFAIK [17:37] gary_poster, yeah, but it did in mine and bcsaller_'s old trunks. [17:37] Makyo, used to be there [17:38] so new builds should never have this error [17:38] and the system is still configured to read that file? [17:38] we could try cleaning out old cruft too [17:38] I suspect that this was some build transition mistake [17:38] Still loaded by modules-debug.js [17:39] Line 116 [17:39] Makyo but built in build- dirs [17:40] Makyo, or served dynamically when in devel mode IIRC [17:40] OK can I get review 2x https://codereview.appspot.com/7312084 [17:40] I can't make my freaking index.html reste [17:40] reset [17:40] * hatch hopes he did it right [17:41] I did a make appcache-force repeatedly, and cleared chrome cache repeatedly, but old index.html still there [17:41] Makyo forgetting anything? Any other ideas ^^ ? [17:41] gary_poster, Yeah, I see that in the lib/templates.js script, but it doesn't jive with what devel's serving, because there's a catchall rule for /juju-ui/*-> app/{file} [17:42] gary_poster, Hmm, just a sec. [17:45] bcsaller_, aha.. re em watch error.. its the inotify limit.. http://stackoverflow.com/questions/10086336/multiple-node-js-always-error-watch-emfile [17:45] its 128 by default it seems [17:45] fixed with echo 8704 | sudo tee /proc/sys/fs/inotify/max_user_instances [17:46] hazmat, bcsaller_: oh man, sorry, I should've looked at your pastebin earlier. I ran into that a while back, too, fixed similarly. Sorry :/ [17:47] Makyo: heh, I increased my limit too already, but I'm still not sure why we are hitting it, maybe we crossed 128 files in the dir or something [17:47] gary_poster, I added a letter and it showed up with make appcache-force, then right click in the "Network" section of the console, clear browser cache. [17:47] trying [17:48] Then it takes a second to reload and the status bar says 'Waiting on AppCache...' [17:48] no luck Makyo [17:49] I'll try restart chromium... :-/ [17:51] nope [17:54] gary_poster, In the network tab, for localhost, does it say (from cache) in the size column? [17:55] Makyo, yes [17:55] and timestamp of manifest is definitely changing [17:56] and I'm definitely clearing cache :-) [17:59] Snagging Chromium. Did you get yours from apt, gary_poster ? [17:59] y Makyo thanks [17:59] on call but will return soon [18:12] Makyo back [18:13] You install chrome from google rather than chromium from apt usually? [18:15] gary_poster, personally, yeah. Just hadn't gotten around to installing chromium yet. [18:15] gary_poster, ctrl+shift+r help any? That made it work on my end. [18:16] changed index, appcache-force, clear browser cache - plain refresh didn't show a change, but ctrl-shift-r did. Different behavior from Chrome. [18:16] no had tried that. will retry. I could also try touching index maybe... [18:18] touching index.html was the trick Makyo . Thanks and sorry to bother [18:18] gary_poster, no worries, good to know! [18:19] hatch, if you link branch to bug that is nice. Can do in command line with bzr commit --fixes=lp:1122706, or in Launchpad from bug or branch. I just linked for you [18:20] It helps people and kanban automation find related artifacts [18:21] ok great, I just made note of that [18:21] am I doing these requests properly? [18:21] they sure take a long time for small changes - are we able to group fixes in a branch? [18:23] hatch, lbox taking a long time? [18:23] hatch, if so, the reason is it's running tests and lint. [18:23] I think he is referring to all of the bug card branch stuff [18:23] hatch you can combine for very small things [18:23] ok :) [18:24] Ah, alright. [18:24] Okay, really need to finish moving stuff outside, back in a few. [18:28] hatch, "land as is" for your branch for me. You need one other person. I would have either combined the HACKING changes into the same branch or just changed it without a card or bug, since it would take no time at all. [18:28] What you are doing is even nicer, to be clear [18:28] but for your own annoyance, I mean [18:29] ok sure thing [18:30] ok finally cleaned up enough ram to boot up this windows 8 vm [18:30] *please don't crash...please don't crash* [18:30] ;) [18:30] anyone here use 'Fitbit' ? [18:31] do we have a Canonical fitbit team? :) [18:31] heh, I don't, heard good things about it though [18:32] it's proof of how lazy I can be some days [18:32] :) [18:32] lazily active? [18:32] actively lazy? [18:33]