| === gary_poster|away is now known as gary_poster | ||
| gary_poster | trunk test pass for me on IE, over three runs :-/ | 01:00 |
|---|---|---|
| gary_poster | the destroy service tests are misbehaving--leaving junk on the page and in the url--but I don't think that's the cause here | 01:01 |
| gary_poster | I don't think that's related to the failures I mean | 01:02 |
| gary_poster | "overwrites config flags with url flags" and "parses url with and without flags" are two of the failing tests. In both cases the error is "TypeError: Object expected" "at Anonymous function (http:.../test/test_feature_flags.js:30:11)" | 01:05 |
| gary_poster | that's the line number for the first error | 01:06 |
| gary_poster | all four look to be the config flag thing | 01:06 |
| gary_poster | I think maybe we can try to use Francesco's new tools to debug these tests | 01:07 |
| gary_poster | maybe the charm is the different variable? :-/ | 01:07 |
| gary_poster | Seems unlikely, but possible | 01:07 |
| gary_poster | rick_h_, ^^ heading out tonight; we can look tomorrow. if you need to pass it to someone else let me know. thank you | 01:08 |
| gary_poster | (same four tests failed twice in a row on saucelabs) | 01:09 |
| === rogpeppe1 is now known as rogpeppe | ||
| frankban | hi rogpeppe: I have a prototype that seems to work: http://pastebin.ubuntu.com/5696777/ | 12:17 |
| frankban | rogpeppe: anyway, I don't know if that's the right direction | 12:17 |
| rogpeppe | frankban: looking | 12:18 |
| frankban | rogpeppe: thanks | 12:19 |
| rogpeppe | frankban: it seems a bit odd that you're passing APIInfo into state.Open | 12:20 |
| rogpeppe | frankban: otherwise it looks good | 12:21 |
| frankban | rogpeppe: yes, but I need to store apiInfo in the state in order to retrieve the api addresses | 12:22 |
| rogpeppe | frankban: hmm | 12:22 |
| rogpeppe | frankban: i'm not sure that's quite the right approach | 12:23 |
| * rogpeppe goes to look at the code | 12:23 | |
| rogpeppe | frankban: i think that rather than passing an Addresser object into NewSimpleContext, it might be best just to pass the addresses directly | 12:26 |
| rogpeppe | frankban: then you'd need to find a way to pass the API addresses into newDeployer | 12:26 |
| rogpeppe | frankban: in cmd/jujud | 12:26 |
| frankban | rogpeppe: indeed, back in a minute | 12:28 |
| rick_h_ | wow, no hatch idling. Did we lose canada last night? | 12:49 |
| gary_poster | heh, I hope not | 12:58 |
| benji | the internet connection between the US and Canada went over the I-5 bridge | 13:03 |
| bac | :) | 13:03 |
| gary_poster | benji, do you have a feel for your schedule today in getting the initial replay work landed? if you have time I'll ask you to do some work on the vitals stuff. I'm kind of slammed today and I think we need progress there. | 13:04 |
| gary_poster | but I don't want to get in the way of your previously planned flight path | 13:04 |
| bac | we did have a local bridge they closed and wanted to tear down until they realized all of the utilities that run under it would have to be re-routed. so they relunctanly made it pedestrian only (what the community had asked for from the beginning) and now it is a big draw. | 13:05 |
| benji | 13:24 | |
| benji | gary_poster: sorry, I just noticed that you spoke to me; I upgraded to R last night and aparently my beep stopped working | 13:28 |
| gary_poster | :-) np | 13:28 |
| benji | re. schedule, I think I can get this done with a couple/three hours left in the day. Will that be enough to help materially? | 13:28 |
| * benji tries to fix the beep | 13:29 | |
| gary_poster | yeah benji thanks. lemme know when you get near the end so we can try to find time to sync up | 13:30 |
| benji | k | 13:30 |
| benji | <sigh> Once again, I have to remove pcspkr from modprobe's blacklist; I think I need to start a benjbuntu distro that will do things my way | 13:32 |
| gary_poster | sinzui, when you have a moment I'd like to talk through luca's goal email to make sure I understand the charm browser parts. would you have time sometime in the next hour? | 13:51 |
| BradCrittenden | jujugui: there are electricians back in the house today causing all sorts of mayhem. i think i've got most of our network on UPS but i may occasionally go offline today. | 13:55 |
| === BradCrittenden is now known as bac | ||
| gary_poster | ack | 13:55 |
| bac | one indicator is me losing my nick... | 13:56 |
| benji | heh | 13:56 |
| bac | so it looks like the overlords of canonistack are very aggressive about deleting old images, even if they are in use. this appears to have deleterious affects on our running servers. | 13:58 |
| gary_poster | :-/ | 13:58 |
| rick_h_ | gary_poster: so the CI is using the normal charm? cs:~juju-gui/precise/juju-gui | 13:59 |
| gary_poster | define "normal" in context, rick_h_ , but yes :-) | 13:59 |
| rick_h_ | gary_poster: I've got my IE10 set and verified tests pass so now looking to deploy the charm as you mentioned that being a difference | 13:59 |
| rick_h_ | gary_poster: k, thx | 13:59 |
| gary_poster | rick_h_, yeah that jibes with my experience | 13:59 |
| hatch | morning | 14:02 |
| rick_h_ | hatch: morn | 14:02 |
| gary_poster | hiya | 14:02 |
| rick_h_ | hatch: can you dbl check the DNS issue from yesterday sometime. Just curious if that's wrapped up. | 14:02 |
| gary_poster | uh oh | 14:03 |
| gary_poster | not | 14:03 |
| gary_poster | not wrapped up | 14:04 |
| gary_poster | only ok if you are on google dns | 14:04 |
| gary_poster | crap | 14:04 |
| rick_h_ | :( was hoping that wasn't the case | 14:04 |
| gary_poster | going to #is | 14:04 |
| hatch | hmm something is wrong then - I've never seen dns take more than 24h | 14:04 |
| rick_h_ | yea, why I wanted to double check. it's into 'something wrong' mode | 14:04 |
| hatch | rick_h_: did you get the sauce lab details? | 14:05 |
| hatch | they are on the CI wiki | 14:05 |
| rick_h_ | hatch: well gary_poster showed me the video point where the tests are | 14:05 |
| rick_h_ | hatch: so working on trying to figure out wtf...tests pass in local IE no problem. Trying to test against a charm serving the tests now in the hope something jumps up and breaks | 14:05 |
| hatch | ugh one of THOSE issues | 14:06 |
| * rick_h_ hates hoping that things break | 14:06 | |
| rick_h_ | hatch: if you've got more insight on "THOSE" make sure you let me have it please :) | 14:06 |
| gary_poster | to get my ducks in a row for is: | 14:06 |
| gary_poster | host jujucharms.com from machine without google dns is 174.36.138.31 | 14:07 |
| gary_poster | good ip is 174.129.219.53 | 14:08 |
| hatch | correct | 14:08 |
| hatch | it's like someone forgot to hit the 'propogate' button :) | 14:09 |
| gary_poster | hazmat you around? the dns switch to jujucharms.com appears to have gone badly, and I wanted to know who to talk to in #is | 14:09 |
| gary_poster | I think you had an RT for this | 14:10 |
| hazmat | gary_poster, RT | 14:17 |
| hazmat | gary_poster, i did one for uistage that went through smoothly to get off hp | 14:17 |
| gary_poster | hey hazmat, talking through this on canonical #is. | 14:18 |
| hazmat | gary_poster, joining there | 14:18 |
| hazmat | gary_poster, dig @ns.canonical.com jujucharms.com | 14:18 |
| benji | hmm, there doens't seem to be an R build of lbox; has anyone else seen that? | 14:24 |
| rick_h_ | benji: yep | 14:24 |
| rick_h_ | benji: had to build it. | 14:25 |
| rick_h_ | ;sudo go get launchpad.net/lbox && sudo go install launchpad.net/lbox | 14:25 |
| rick_h_ | benji: ^^ well minus the ; in my lxc got it going. | 14:25 |
| benji | thanks! | 14:25 |
| rick_h_ | man, this just sucks. Chrome/FF won't even run the tests on the charm over https. Ok, should be an easy fix. IE passes no problem, just a warning on mixed content. FF fails on local test run. /me is so confused | 14:27 |
| gary_poster | rick_h_, charm lets you switch to insecure mode for just this reason | 14:28 |
| gary_poster | but doesn't help the issue | 14:29 |
| rick_h_ | gary_poster: oh, missed this config. /me goes back to the config | 14:29 |
| rick_h_ | gary_poster: yea, pushing a trivial change to fix the https issue I think. | 14:29 |
| gary_poster | cool thanks | 14:29 |
| rick_h_ | gary_poster: but still IE works against the charm (ugh) and FF fails locally (double ugh) | 14:29 |
| hatch | I wonder if any of this is related to mocha | 14:31 |
| hatch | I don't really know how mocha works internally | 14:31 |
| gary_poster | rick_h_, FF fails locally for me too. frankban said it was probably because FF was not in focus | 14:32 |
| hatch | blame what you don't understand! | 14:32 |
| hatch | ;) | 14:32 |
| gary_poster | FF must be in focus for tests to pass apparently | 14:32 |
| gary_poster | have not rerun to verify since then | 14:32 |
| rick_h_ | gary_poster: ok. Yea, I recall that issue coming up. So I can not worry about FF. So my only real problem is that IE passes against the charm and CI doesn't. | 14:34 |
| gary_poster | right :-( | 14:34 |
| gary_poster | and the error is not particularly helpful | 14:35 |
| gary_poster | rick_h_, one approach would be to do what you want and make it a yui module and test it that way :-P | 14:35 |
| rick_h_ | no, IE has some lovely errors. The lines in the tests that's against doesn't fit the error really | 14:35 |
| hatch | you can join the ci test in progress and open the dev tools in IE | 14:36 |
| rick_h_ | hatch: yea, looks like that'll have to be the technique to move it forward. I was wondering if maybe it's some quirks mode/etc that IE might be in but can't replicate the failures locally | 14:37 |
| rick_h_ | hatch: are there docs on how to do that? | 14:37 |
| hatch | when the test starts click the link | 14:38 |
| hatch | then click the window to have the vm avail to you in the browser | 14:38 |
| hatch | ^^ docs ;) | 14:38 |
| rick_h_ | :) | 14:39 |
| hatch | haha sorry I don't remember if that specific usecase was documented | 14:39 |
| rick_h_ | hatch: all good, thanks | 14:39 |
| hatch | http://simonsfoundation.org/features/science-news/unheralded-mathematician-bridges-the-prime-gap/ | 14:39 |
| hatch | interesting read for anyone who is interested in mathy things | 14:40 |
| sinzui | hi gary_poster. I am now free to discuss jujucharms.com | 14:45 |
| gary_poster | sinzui, ok, thanks. Lemme finish up in #is and then I will meet you in 4 minutes in guichat for 10 minutes? | 14:46 |
| sinzui | fab | 14:46 |
| teknico | lbox/rietveld extremely slow today... | 14:47 |
| rick_h_ | teknico: might be LP as well. I know some days LP is slow it drags things along. | 14:47 |
| gary_poster | jujugui call in 9 | 14:52 |
| gary_poster | pls update kanban | 14:52 |
| hazmat | gary_poster, re resolving locally for dev ... using an external dns server like google's public dns is an option | 14:52 |
| gary_poster | hazmat, ack, yeah recommended that on blog which made it through to juju blog | 14:53 |
| gary_poster | jujugui call in 1 | 14:59 |
| benji | gary_poster: installing plugin... | 14:59 |
| rick_h_ | yay I blew things up! | 15:18 |
| rick_h_ | hatch: when you get a sec would love a steal a few min of your time | 15:22 |
| hatch | sure going to be in the weekly call for a bit | 15:22 |
| rick_h_ | hatch: rgr | 15:22 |
| rick_h_ | victory! sorta | 15:37 |
| rick_h_ | ok, gary_poster and hatch would like to chat post weekly. I've got this but bigger issue in there. | 15:41 |
| hatch | alrighty | 15:42 |
| bac | gary_poster: new branch is just simply lp:charms/juju-gui | 15:58 |
| gary_poster | rick_h_, hatch guichat now? | 16:04 |
| hatch | sure | 16:04 |
| rick_h_ | gary_poster: rgr | 16:04 |
| rick_h_ | hatch: gary_poster posted bug #1183887 please let me know if something isn't clear per our discussion or edit it a bit | 16:36 |
| _mup_ | Bug #1183887: mocha silently skips bad test files <tests> <juju-gui:Triaged> <https://launchpad.net/bugs/1183887> | 16:36 |
| gary_poster | perfect, thanks rick_h_ ! | 16:36 |
| hatch | thanks rick_h_ | 16:47 |
| hatch | I also think I fixed my internet | 16:47 |
| hatch | looks like the isp auto updated my router reverting some settings to their defaults | 16:48 |
| hatch | and CLEARLY I need more power ;) | 16:48 |
| rick_h_ | orangesquad running off to lunch before CI tells me anything else it doesn't like about my stuff. bbiab. | 16:51 |
| sinzui | thanks rick_h_ | 16:51 |
| hatch | lol | 16:59 |
| * hatch grabbing some lunch | 17:26 | |
| gary_poster | hatch we have our call scheduled now :-) | 17:32 |
| * gary_poster decides hatch isn't coming and go gets lunch himself, in a 12 minute slot suddenly available :-) | 17:48 | |
| rick_h_ | orangesquad I'm back alert | 17:56 |
| rick_h_ | yay, the build is back to normal. /me does a happy dance | 17:58 |
| hatch | gary_poster: oh crap so sorry! I totally missed it on the calendar | 17:59 |
| hatch | woudl you like to reschedule? :) | 17:59 |
| gary_poster | hatch np. i'll ping you if I have time this afternoon, or we can do it next week | 17:59 |
| hatch | sure thing | 18:00 |
| hatch | neighbour down the street just bought a new rav 4 - damn they look nice | 18:02 |
| rick_h_ | subaru or bust! | 18:02 |
| hatch | haha | 18:03 |
| hatch | I was contemplating selling the 4runner for an STI | 18:03 |
| hatch | but they are just too expensive here | 18:03 |
| rick_h_ | outbacks rule. Wife's driving my old one now. Love those things. | 18:04 |
| hatch | do you have the one with the mini truckbox? | 18:04 |
| hatch | or is that not an outback | 18:04 |
| rick_h_ | no, crosstrack or something | 18:05 |
| hatch | ohh | 18:05 |
| rick_h_ | I'm a fan of the original. wagon, drive through anything, haul 8' crap, get 25+mpg outback | 18:06 |
| rick_h_ | rav4 is a can of meh in comparison :P | 18:07 |
| hatch | haha | 18:07 |
| hatch | I dont think I could actualy not have the 4runner | 18:07 |
| hatch | need to pull things :) | 18:07 |
| rick_h_ | that's why the wife drives the outback now. I got my brake controller installed and went camper shopping over my vacation | 18:08 |
| hatch | nice - pick anything up? | 18:08 |
| rick_h_ | not yet. I promised I'd get the touareg this year and a camper next year. but liking the starcraft 227cks a lot and might get that | 18:09 |
| hatch | cool - have you noticed how the interiors of the new campers look like they are from the 80s? | 18:10 |
| hatch | haha | 18:10 |
| rick_h_ | heh | 18:10 |
| hatch | I suppose that makes them more timeless | 18:11 |
| bac | hatch: here is a story that is the exact opposite of the number theory episode you mentioned: http://projectwordsworth.com/the-paradox-of-the-proof/ | 18:12 |
| hatch | cool I'll read it later - can you give me a tldr? | 18:12 |
| hatch | rick_h_: there is a bug which causes charms deployed from the browser to not have all of the config data - I am guessing this has to do with the charm.loaded = true; fudge. | 18:17 |
| hatch | can I simply make a request to pull in the missing details? or do you have a technique for this already? | 18:17 |
| rick_h_ | hatch: I've not looked yet. I thought we already worked around the loaded = true stuff | 18:18 |
| rick_h_ | so something changed that broke it and not sure what's changed around there. | 18:18 |
| rick_h_ | seen the bug, not looked for the source of it yet | 18:18 |
| hatch | umm it was in the trunk that I branched earlier in the week | 18:18 |
| hatch | ok let me rephrase... | 18:18 |
| rick_h_ | hatch: yea, but that was working when the feature was released. benji actually did it before 13.04 | 18:18 |
| hatch | when the charm is 'added' should all the data be there already? or do we only pull in a minimum amount? | 18:19 |
| rick_h_ | hatch: it should be everything. It pass in an instance of BrowserCharmModel and create an instance of CharmModel sent along | 18:19 |
| hatch | ok because I remember a discussion we had in which we talked about only sending the minimum data | 18:20 |
| hatch | I'll investigate further | 18:20 |
| hatch | thanks | 18:20 |
| rick_h_ | hatch: yea, let me know if you need a hand. I helped benji some so know how it kind of fit together. | 18:20 |
| rick_h_ | hatch: just lower down my stack of todo atm | 18:20 |
| hatch | yeah it's alright it's getting close to your EOD on a long weekend :) | 18:21 |
| rick_h_ | woot! Need it, my head feels run through the wringer today. | 18:22 |
| hatch | haha - I gota find some time to get the kennels to the.....kennel, for next week :) | 18:23 |
| hatch | I'd love it if I could connect my computer to my home internet and my two mobile connections to download things at 3x the speed :) | 18:37 |
| hatch | I'm sure it could be done with 3 wifi chips and some torrent software | 18:37 |
| hatch | rick_h_: still around? | 18:39 |
| rick_h_ | hatch: yep | 18:39 |
| hatch | ok the charm config template is looking for a 'settings' property which is absent from charms deployed from the charm browser | 18:39 |
| hatch | it looks like they are now called 'options' | 18:40 |
| hatch | or am I missing something? | 18:40 |
| rick_h_ | hatch: looking, sec | 18:40 |
| rick_h_ | hatch: guichat? | 18:43 |
| hatch | sure | 18:43 |
| jcsackett | rick_h_: was there something besides test-server.sh that needed the s/localhost/0.0.0.0/ treatment? | 18:45 |
| benji | gary_poster: combining the fact that I forgot about the weekly meeting and that this refactoring/testing is taking longer than I expected, I don't think I will have much time left over today; I'll let you know when I am done so I can do what I can | 18:48 |
| gary_poster | ok benji np. thanks for heads up | 18:48 |
| rick_h_ | jcsackett: I think that's the one. jcsackett maybe the rapi stuff as well? | 18:53 |
| rick_h_ | jcsackett: but we don't tend to have to have that with the sandbox any more so meh | 18:53 |
| jcsackett | rick_h_: ok, i'll check that later. i was mostly asking to see if i could just do it as a driveby in my current branch, or if it was worth spinning up a separate branch for it. i'll go with driveby. | 18:53 |
| * hatch feels better now that he wasn't the only one who forgot about the weekly call :) | 18:53 | |
| rick_h_ | jcsackett: rgr | 18:54 |
| hatch | gary_poster: is it looking like you'd like to do the chat today or push to next week? | 18:56 |
| rick_h_ | hatch: so settings appears to come from app/views/charm-panel.js which is config && utils.extractServiceSettings | 18:56 |
| gary_poster | hatch, in 1 hr? | 18:56 |
| rick_h_ | hatch: so I think just getting options from browsercharm into charm.config should work out for you | 18:56 |
| hatch | rick_h_: yep just saw that | 18:56 |
| hatch | :) thanks | 18:56 |
| rick_h_ | hatch: no problem, always here to help point out the obvious :) | 18:56 |
| hatch | gary_poster: sure - I'll take this time to quickly run and drop off the kennels now | 18:56 |
| gary_poster | cool hatch ttyl | 18:57 |
| rick_h_ | hatch: so you're working on that bug then? I'll pull our card off the board for that if so. | 18:57 |
| hatch | rick_h_: umm....well yes I'll resolve it | 18:57 |
| rick_h_ | hatch: k, thanks. Make sure to link to the bug in the branch/etc. #1182295 | 18:57 |
| _mup_ | Bug #1182295: adding a charm from the charm browser is broken <blocker> <charmbrowser> <juju-gui:Triaged> <https://launchpad.net/bugs/1182295> | 18:57 |
| hatch | although this is only an intermediary branch until we get the new UI | 18:58 |
| hatch | ok will do | 18:58 |
| sinzui | jcsackett, I see you are stabbing localhost rules in the gui | 19:12 |
| jcsackett | sinzui: yup. | 19:12 |
| Makyo | gary_poster, ping | 19:36 |
| gary_poster | Makyo, thanks sorry finishing with ben | 19:36 |
| Makyo | gary_poster, np. | 19:36 |
| hatch | oh boy the traffic was nuts | 19:46 |
| Makyo | gary_poster, http://bazaar.launchpad.net/~juju/juju-core/trunk/view/head:/cmd/juju/upgradecharm.go has the help-text for the options | 20:08 |
| gary_poster | awesome thanks Makyo | 20:08 |
| hatch | ugh | 20:08 |
| hatch | now manage.jujucharms.com is cutting off the transfer part way through | 20:09 |
| gary_poster | hatch, of what? http://uistage.jujucharms.com:8080/bws/sidebar/worked for me | 20:10 |
| hatch | https://manage.jujucharms.com/api/0/charms/interesting | 20:11 |
| hatch | Uncaught SyntaxError: Unexpected end of input | 20:11 |
| hatch | but only when opened through the app | 20:11 |
| hatch | direct it appears to return properly | 20:11 |
| hatch | albeit with one HUGE response heh | 20:12 |
| gary_poster | yeah, I think it will be good to serve the icons separately. that is still the plan right rick_h_ ? | 20:12 |
| hatch | yeah no matter what I do the response is cut short | 20:13 |
| gary_poster | hatch, I get interesting on left | 20:13 |
| gary_poster | and I can expand ceph on right | 20:13 |
| hatch | yeah it DID work | 20:13 |
| hatch | hmm | 20:13 |
| hatch | doesn't work on uistage for me either | 20:14 |
| hatch | maybe my internet does hate me today | 20:14 |
| gary_poster | :-) | 20:14 |
| hatch | I vote to write a local api simulator for jujucharms :) | 20:15 |
| hatch | oh hey it worked! | 20:15 |
| hatch | and there it goes | 20:15 |
| hatch | lol | 20:15 |
| hatch | this is pretty cool how you can use the 'old' panel and the 'new' one with the flags | 20:19 |
| gary_poster | hatch you available for call? | 20:24 |
| hatch | sure thing | 20:24 |
| sinzui | abentley, do you have a moment to advise me about the direction to fix https://bugs.launchpad.net/charmworld/+bug/1180055 | 20:35 |
| _mup_ | Bug #1180055: KeyError: 'short_url' <oops> <charmworld:In Progress by sinzui> <https://launchpad.net/bugs/1180055> | 20:35 |
| abentley | sinzui: Sure. | 20:35 |
| sinzui | ^ I can ensure the charm has an empty short_url, or I can just update the tal to check for the short_url before attempting to make the tag | 20:36 |
| sinzui | but maybe you feel some earlier process should have created the short_url. | 20:36 |
| abentley | What is the short_url used for? I didn't need to include it in the API. | 20:37 |
| sinzui | well, some irony is this case. | 20:38 |
| sinzui | abentley, /tools/store-missing view wants to link to the charm in m.jc so that you can view it. but not every charm has a short url to make that link | 20:39 |
| sinzui | abentley, My quick tal hack shows let me see 3 of the problem charms, but the first two don't have pages | 20:40 |
| sinzui | abentley, I think http://staging.jujucharms.com/tools/store-missing want to show me a link to http://staging.jujucharms.com/~brunopereira81/oneiric/teamspeak3 which works. the juju-gui link in the same list does not work | 20:41 |
| abentley | sinzui: I think every charm should have a short_url. We have all the info needed once we've called available_charms. | 20:43 |
| abentley | sinzui: It would be a trivial addition to abel's branch. | 20:44 |
| sinzui | The search results implies that. I can follow the link to /charms/precise/juju-gui, then m.jc decides there is a problem and redirects to the MIA listing, and it dies on the URL that I can easily construct from the available info | 20:45 |
| abentley | sinzui: Does that clear it up for you? | 20:47 |
| sinzui | abentley, since the view is redirecting to MIA, do we want to make a link back to problem charm? the user will be in a cyclic loop | 20:47 |
| abentley | sinzui: No, it doesn't seem productive to have a link to a charm known to be inaccessible. | 20:48 |
| abentley | sinzui: Could you please review https://code.launchpad.net/~abentley/charmworld/api-docs/+merge/165696 ? | 20:50 |
| sinzui | sure | 20:51 |
| sinzui | r=me abentley | 20:56 |
| rick_h_ | gary_poster: yes we'll have the icons out of the api shortly | 21:15 |
| gary_poster | yay rick_h_ thanks | 21:15 |
| rick_h_ | it's up after getting the browser behind the FF | 21:15 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!