[00:18] popey: BUUUUUG === lifeless_ is now known as lifeless === zumbi is now known as Guest40451 [02:46] Night guys === smb` is now known as smb [08:38] *yawn* [08:38] morning, all [09:36] morning all! [10:07] Good morning all! :D === Chipaca_ is now known as Chipaca [11:31] buenos días! [11:45] nessita, morning! [11:46] nessita, I have an interesting question for you, got time? [11:46] mandel: sure! [11:47] nessita, I've been looking at writing integration tests with squid + ssl so that we can test the cases in which we have a selfsigned and get an error [11:47] nessita, what happens is that due to license issues the package of squid3 that ubuntu ships does not use the --enable-ssl flag, which means that it does not support that [11:47] * Daviey wonders if it turned out not infact to be interesting, if nessita could redact her offer of time. [11:48] nessita, with some help from #ubuntu-servers I've got a ppa with squid3 + ssl for the tests.. but, how do we manage this? [11:48] nessita, should I forget squid and look into a diff path? mock the ssl issues for example? [11:49] mandel: guh... question, what --enable-ssl provides? [11:49] mandel: Do you need a full proxy, or would a stub https server be enough? [11:49] nessita, http://www1.it.squid-cache.org/Versions/v3/3.HEAD/cfgman/https_port.html [11:49] Daviey, at the moment we are using a full proxy to try and have the integration tests as real as possible, if not I think I can do mock [11:50] nessita, I can also do squid3 -v check if the flag was used and skip the ssl tests if it wasn't [11:50] mandel: it seems to me that you only need to test as far as the handshake.. [11:50] we simply cannot enable redistribution of squid linking with openssl. [11:51] Daviey, I know.. licenses.. [11:52] mandel: would you please tell me what does --enable-ssl provides (without using a link for the answer)? :-D [11:52] nessita, ppa:lynxman/squid-ssl already compiled for amd64 http://people.canonical.com/~lynxman/squid3/ [11:52] mandel: let's go step by step, I'm trying to understand your issue before understanding one of the solutions (the PPA) [11:52] nessita, it allows squid to listen connections over https [11:52] mandel: if you want to add gnutls support for squid, i'd be over the moon :) [11:53] mandel: ok, so, without that, you can not set a proxy test to be https://localhost:foo [11:53] Daviey, no no no, I don't have the time to fix the world :) [11:53] nessita, yes, that is the issue [11:53] * nessita thinks [11:54] mandel: if you have the patch file that adds that support, we could add that to our PPA, though I'm not thrilled to do that [11:54] Daviey: hola! [11:55] nessita, yes, I'm not happy either.. [11:55] Daviey: what's the licensing issue with that part of squid? [11:58] Daviey: ping? [12:01] nessita: sorry, it's BSD and GPL linking [12:01] YOU can do it, WE just can't redistribute it. [12:03] mandel: I guess a pacthed squid in our ppa can work. We just need the patch to apply, and we can upload the new squid to ppa:ubuntuone/nightlies [12:03] nessita, is that an ok solution, I mean, I don't want to screw people because of this [12:04] mandel: is not a problem, as far as I can see, we can check with dobey if he sees any issue, but I don't see one [12:04] Daviey: thanks! [12:04] nessita, cool :D [12:06] mandel: I read your comments to my review. ABout the fakes... why you need to fake the arrow? why not juts letting the code use the real one (since is there in the pythonpath)? [12:08] nessita, I don't mind using the real one, yet a prefer tests not to relay in real code, I mean, is the arrow is broken, the arrow tests should fail but not the expander ones, but if you prefer not to do that I'm happy to remove it :) [12:08] nessita, I also misunderstood what you meant with the expanded arg, fixed that, you where right, I called the set_expanded with it rather than store it as false [12:08] nessita, and removed the @property to use property() to make the code cleaner [12:09] mandel: so, you're right about that when it comes to 100% pure unit tests. But I personally think that if we *only* build 100% unit tests, we're missing a lot of things. And since we don't have the time to write 100% unit tests and thin layers of integrating modules, I advice to have a test suite that runs relatively fast, and to mock only the strictly necessary [12:10] mandel: so, unless something is a pain (such as using a socket, for example, or an unpatchable class, or sys.modules, or something like that) I would advice not doing mocking [12:10] nessita, sure, no problem, what ever is best for the team, I'll remove those, but.. any ideas on how to assert that the correct methods have been called? or should I just test that the get returns the expected value [12:10] ?? [12:11] mandel: the latter, since testing that methods 'a', 'b', 'c' were called in that order will break the test when we change the method 'a' to be 'z' which does the same as 'a' but is better, for example [12:11] o/ [12:12] nessita, ok, should be a simple change, let me get down to it [12:12] mandel: or sometimes you test 'a', 'b', 'c' were called, and the we move some stuff around in our internal implementation due to refactoring, and that will break the test (and, depending on the case, shouldn't) [12:12] hola duanedesign [12:13] nessita, I think I was just stating that the method was called with the args, but you are right, I'll just assert that the expected value is present [12:13] nessita, and is easier to remove lines than to add them :) [12:18] mandel: right. Ack to the setter and _expanded prop [13:26] alecu: hola! I need some help from your end [13:27] alecu: would you have some minutes? [13:33] good late morning! [13:36] Daviey: because squid is BSD and openssl is GPL? [13:39] Hey, JAVA devs out there - We're in the review stage of Ubuntu One Files Java Client, which is a library that allows to talk to U1, manipulate files, up/download, etc. If you'd like to give it a go, this is the last branch: [13:39] lp:~karni/ubuntuone-files-java-client/interactive-cli-example [13:39] If you like U1 and code for Android, you may want to find it of interest, too :) [13:41] dobey: other way around, but openssl is bsd-like [13:41] ok, I'm off to have lunch [13:41] * mandel lunch [13:43] Daviey: because openssl has advert clause then? [13:44] alecu: ping? [13:45] dobey: http://www.squid-cache.org/mail-archive/squid-dev/200406/0011.html [13:46] aquarius: If anyone asks about U1 Java library, I just shouted out we are happy to have initial public preview, as stated few lines above. I'd like to weed out any rough edges with whoever uses it first (including me ;)) [13:46] joshuahoover: ↑ [13:46] karni, cool! [13:46] mandel, dobey, Daviey: that probably means noone ever uses squid with ssl, so we canjust skip those tests, I say [13:47] ralsina: squid with ssl is a way to have a proxy for https, which a lot of people may have (even though is not squid behind that) [13:47] karni, we should chat with roberta about it; she's at MWC this week though [13:47] ralsina: so we need that to build the integration tests [13:48] nessita: yes, but we are not going to work around this by packaging an illegal squid for ourselves, are we? [13:48] aquarius: Agreed. Although I prefer to treat it like u1db 'preview' - once we know it's prime time ready, we can announce it. [13:48] ralsina: is not an illegal squid, as far as I understood it [13:48] nessita: it's not distributable [13:48] aquarius: I have a dead simple CLI example, and will probably quickly write a UI-enabled example. [13:49] nessita: but I may have missed some licensing subtlety [13:49] nessita: we would have to each build it ourselves locally [13:49] nessita: it couldn't be in the PPA [13:49] dobey: ok, so we can build it in those system running the squid tests, no? [13:49] karni, definitely, that's why I want to talk to roberta :) [13:49] well, unless we used gnutls instead [13:49] aquarius: ah, cool :) [13:49] karni, and you and I should chat about it; a blog post by you would be cool, I think [13:50] aquarius: That would be nice :) [13:51] alecu: you around? [13:52] nessita: i'm not sure testing that works against squid, will tell us anything useful [13:53] aquarius: Also, this is a silly detail, but I've grown to think ubuntuone-java-files-*client* is not a great name for a library [13:53] aquarius: What do you think? [13:54] dobey: why not? [13:54] karni, well... it isn't, really, but I don't know how Android and Java stuff like this is really named [13:54] aquarius: Same goes for ubuntu-sso-java-*client* - it is a client of the API, but a library to the software that will use it [13:54] aquarius: say, google analytics builds a libGoogleAnalytics.jar [13:55] karni: +1 on removing the client [13:55] ralsina: ack [13:55] karni: if anyone tries to use it as a server, he will figure it out soon enough ;-) [13:55] aquarius: none of the libraries we use in the u1-files-java-client (we use like 13) have "lib" or "library" in their names [13:55] ralsina: haha [13:55] nessita: because we're testing against a configuration which noone should actually have anyway. there's no guarantee that squid and IIS and whatever else, all do ssl proxies the same way. we need to test against actual server configurations people will be using [13:56] karni, do java things normally get called a "library"? It'd be useful to see what similar stuff is called [13:56] nessita: also, have you seen https://bugs.launchpad.net/ubuntu/+source/ubuntuone-client/+bug/942312 ? [13:56] aquarius: I'll google [13:57] aquarius: not really. they just call by names, like commons-logging-1.1.1.jar, jackson-mini-1.9.4.jar, signpost-commonshttp4-1.2.1.1.jar [13:58] aquarius: things that are libraries do; though people generally refer to them by name as in any other language [13:58] dobey, this is my point: I'm not sure the Java world talks about "libraries" in the same way we do [13:59] aquarius: what way do we talk about libraries? [13:59] dobey: regarding the first comment, I agree, though since having real configs is very hard (for now at least, on every day testing envs), I would like to have at least one way to test the SSL cert thing [14:00] dobey, a block of reusable code, packaged up on its own with the intention of being included in another project wholesale, is called a "library"; in C it's normally called libsomething. Java... I'm not sure they'd call that a library. I think they might call it a jar, for example. [14:00] dobey: yes to the second, was pinging alecu about it [14:01] aquarius: in java, everything is a jar. in C/C++ we only name the binaries libfoo.so out of tradition. and to make the compiler command line look nicer. there really isn't any technical requirement that it be that way. in fact, the correct technical term for anything with an .so in *nix is "shared object" [14:02] dobey, exactly my point. If a Java person is looking for a shared object to help them talk to U1, would they google for "u1 java library", or something else? I don't know, which is why I asked karni :) [14:03] * karni reads, was googling around [14:03] aquarius: they would probably search for "ubuntu one java" and see what comes up. they might search also for "class library" or "implementation" or other technically useful terms [14:03] aquarius: i don't ever search for "library" really when i am looking for some implementation in C or python or whatever [14:04] karni, call it jar jar binks! [14:04] aquarius: it's a term that is a bit too general and will often dilute results [14:04] mandel: heh :) [14:05] aquarius: they both use library or package, library to do X or package to do X [14:05] aquarius: dobey: I also stumbled upon this http://www.debian.org/doc/packaging-manuals/java-policy/x104.html [14:06] and this http://stackoverflow.com/questions/4177762/naming-java-library-how-to [14:06] Candidates would be: jUbuntuOneFiles.jar , libUbuntuOneFiles.jar [14:07] like jMock.jar, libGoogleAnalytics.jar [14:08] libUbuntuOneFiles.jar sounds sensible to me [14:09] aquarius: The library jar name is not such a problem, the project name is. Clearly, as I've stated, I have mixed feelings towards "Ubuntu One Files Java Client" - if we drop client, it just sounds funny :) [14:09] nessita: what do we need to test with the "ssl cert thing"? we can probably fake it, if it's just to test the dialog and parsing or something? [14:09] aquarius: maybe the project name could contain 'api' keyword, then when we publish Java music library, it have a consistent name [14:09] well I wouldn't be surprised by ubuntuone-files-client.jar either [14:09] aquarius: The project name, that is. [14:10] "Ubuntu One Files Java library" seems OK to me? [14:10] karni: what's the package name: com.ubuntu.one....? [14:10] dobey: we have those (unit) tests in place, mandel was hoping to also have integration tests like with the rest of the proxy support work [14:10] karni: are you just wanting to rename the project? [14:10] pedronis: That's also the question. Currently it's com.ubuntu.one.api.files [14:10] pedronis: We're not sure if it should be com.ubuntuone.api.files [14:11] nessita: and what would it do? ensure that squid serves an SSL cert to us? [14:11] karni: I think one.ubuntu.com is the more official url, but not sure [14:11] pedronis: Our U1F Android app is com.ubuntuone.android.files, SSO lib is com.ubuntu.sso [14:11] dobey: no, to make squid returns an expired cert, or a self signed cert [14:12] karni: ah so we used com.ubuntuone for something already, better stick to it then [14:12] we own both anyway [14:12] dobey: I'm looking for most sensible name for the project name - I think we have just found a nice name for the jar files. [14:12] nessita: can we not just shove those types of certs directly into the tests without squid? [14:13] pedronis: nessita: mandel: Do desktop apps use 'ubuntuone' in their package name? (As opposed to com.ubuntu.one...) [14:13] karni: yes [14:13] In that case we should stick with ubuntuone it seems. [14:13] ubuntuone-client, ubuntu-sso-client, ubuntuone-control-panel, libubuntuone, rhythmbox-ubuntuone, ubuntuone-storage-protocol [14:14] dobey: Oh, thank you [14:14] dobey: as far as I understand, no, I think we need to have squid returning what we need for the test. I guess we could patch what cert to return, but I'm not sure we can patch squid so it handled https url using the whole ssl "stack" [14:14] Album Savings: $57.20 compared to buying all songs [14:14] nice [14:15] guys after opening ubuntuone-control-panel-gtk i am getting: File Sync error (org.freedesktop.DBUs.Error.NoReply: Did not receive a reply .. any idea from where I can start debugging please? [14:15] nessita: i'm not sure why squid has to do anything here. [14:16] Yancho_: ~/.cache/ubuntuone/log/syncdaemon-exceptions.log is probably the first place to look. other log files in same directory would be second [14:16] Thanks for the brainstorm everyone! [14:17] dobey: as far as I know, we have the test suite X, that we run without having a proxy, and with a proxy setup using squid. The same test suite X has to pass in both scenarios. We also have some specific test suite to be run when squid is setup, and those test suite depend on the "proxy" to have specific settings/values. mandel is trying to add a new suite where the "proxy" is an https url, and thus with that he expects squid to serve all the ssl bits a [14:17] mandel: correct me if I misunderstand something [14:17] tnx dobey [14:18] nessita: i understand what is trying to be done. what i don't understand is *why* it has to be that way. [14:18] dobey: can you please repharse your counterproposal then? [14:19] nessita: why do we need integration tests with squid ssl support, when squid with ssl support is so unlikely to be the real world situation? [14:20] dobey: we need a test case with a setup where the proxy is an https url and the proxy will give ceratin ssl certs to perform different tests. We're doing it with squid since we already have the testscases to do it, but we could do it with another proxy setup (no idea which one though) [14:20] nessita: *WHY* ? :) [14:21] dobey: because we need integration tests for the https proxy [14:22] dobey: how would you do that ^, without having squid configured to provide a https proxy? [14:22] we need integration tests with real world proxies. i don't think we need those integration tests to be in the normal test suite, but rather something we run in jenkins against actual real-world proxy servers also set up somewhere [14:23] dobey: I think that's a separated discussion (and it makes sense, but is out of the scope of this talk IMHO) [14:23] nessita: for now, the way to test that is probably through manual testing [14:23] dobey: if we can have some sort of automation for this, we should do it. Manual testing does not scale... [14:23] nessita: then the scope of this talk is squid, and i think we don't need to have integration tests for something nobody is using. [14:24] nessita: yes, which is why i said it should be in jenkins, and we should have manual tests being run by qa or whoever, until we can automate it [14:25] dobey: the fact that squid is behind those tests is "coincidence" (in the sense that is the thing we can setup automatically). If you can propose a way to have another proxy setup differently, let's talk about that. And I propose this separated from the fact that we should run the proxy suite for each possible proxy setup we want to support. [14:26] ralsina: your thoughts? ^ [14:26] i think we should not do integration tests at the moment then, as we haven't the means to do them. [14:26] nessita: I did some minor research right now, and there is no free proxy we can use for this [14:26] ralsina: besides squid, you mean? [14:26] and we can't block on it. So I propose we find a way to set one of these on jenkins [14:27] nessita: I propose setting up a ssl-squid in jenkins and do a job running the tests there [14:27] ralsina: yes, that's my proposal as well [14:27] nessita: but we can't distribute the squid-ssl package [14:27] ack to that [14:27] nessita: ok then, we agree ;-) [14:28] ralsina: agree with dobey now :-P [14:28] nessita: OTOH, we can't get blocked with this [14:28] THERE ;-) [14:29] dobey: the point of squid-ssl tests is that there are certain things we can test there, like whether we are rejecting invalid certificates and so on [14:29] dobey: I have no idea whatsoever on how to fake that, unless we recorded a session or something like that [14:30] ralsina: that seems like something that belongs in a unit test, and not an integration test [14:30] ralsina: one of the goals is not to fake that (just FYI) [14:30] dobey: it can't be a unit test if it depends on the behaviour of other software. That's the whole point of integratiokn tests. [14:31] nessita: I know, I was the one that pushed for integration tests for proxys in the first place :-) [14:32] ralsina: the problem is that tests depending on other software, that nobody is using, isn't useful though. and i feel it's a waste of time to do all the work to make it work for squid, and then we say it works and ship it, and ssl proxies doesn't work for anyone in the real world. [14:32] ralsina: we effectively have the same statistics as only testing the unit side of it. [14:32] dobey: no, because the protocol for SSL proxys is a standardprotocol [14:33] dobey: it's like saying that testing agains IIS is not testing HTTP [14:33] ralsina: IMAP is a standard too. but we all know how well that worked out. [14:33] ralsina: and it would be true. :) [14:33] dobey: yes. This may be more like HTTP than like IMAP, though [14:33] you will be surprised then i guess :) [14:35] dobey: ha [14:36] So, nessita, let's think about getting this on jenkins === m_conley_away is now known as m_conley [14:37] i'd recommend poking the platform/server/debian people to ship squid built against gnutls [14:37] dobey: well, that's a great solution to have thought about 3 months ago ;-) [14:38] dobey: does it even *have* a gnutls build option? [14:39] dobey: mandel poke Daviey and apparently they are not doing it... [14:40] nessita: whenyou have a slot, I changed the spacing in the argparse thing. (damn, I was so close to get that one through without a needsfixing ;-) [14:40] ralsina: --enable-ssl; gnutls has compatibility stuff that replaces openssl headers and uses gnutls compat lib instead. don't know if it "just works" though [14:41] ralsina: ack! [14:43] nessita: as i understand it, it's just that openssl won't (can't) be done. [14:43] dobey: sorry, I already contextswitched my brain. Would please give me more context for that? [14:44] let's wait for mandel [14:49] ugh, i think sinuses are making my lungs hurt [14:52] dobey, I'm back, let read the backlog [14:57] dobey, ralsina, nessita so, yes, the ssl enabled squid stuff is a very strange scenario.. but we don't have numbers about that. If it was not used they wouln'd provide the --enable-ssl flag [14:57] also, we cannot distribute it, so its a matter of having that in jenkins [14:58] what I propose if to have the SquidTestCase to check if the --enable-ssl flag was used, and if it was enable the https ports, otherwise skip it [14:58] mandel: where did you actually discuss the license issues outside of #ubuntuone? did you discuss the possibility of building it with gnutls instead? [14:59] dobey, it was mentioned.. but was just a proposal. I have not looked into that. I did most of my talks in private with one of our guys from ubuntu-server [15:00] dobey, there is a bug about that with a won't fix due to license issues (open-ssl) version, I did not look more into using gnutls [15:00] dobey, mandel: for now, I would discourage trying to build with gnutls, giving everyone's time constraints [15:00] me [15:00] nessita, +10000000 ** 100 [15:00] me [15:02] dobey, ralsina, alecu, briancurtin, standup? [15:02] me [15:02] not me [15:04] me [15:04] me [15:05] nessita, remember we have the mac boy ( urbanape) with us :) [15:08] mandel, after the standup, can you give me a summary update on the SSL issue? [15:09] alecu, sure, is an easy one :) [15:09] nessita: are you going to go with the standup, or we will wait forever for ralsina ? [15:10] mandel, (btw: I'm adding some SSL integration tests to the SD proxy stuff, so I'm adding a dummy certificate and key and some code to set that up) [15:10] please go ahead [15:10] while not ralsina: time.sleep(1) [15:10] DONE: more on bug #933576, reviews, holiday on Monday [15:10] TODO: the same! :-) also, start conversation about Control Panel's new look on Ubuntu [15:10] BLOCKED: nopes [15:10] NEXT: mandel [15:10] DONE: looked at setting tests to use https with squid. --enable-ssl is not there, we need to agree (I think we already did) in the approach. Removed most of the fakes from my ss-dialog branch. [15:10] TODO: Remove the rest of the fakes, ping nessita +1 to review it. Implement what we agree on about squid + ssl. [15:10] BLOCKED: not more than usual. [15:10] next: briancurtin [15:10] * ralsina is currently on his 112th minute on the phone, thanks [15:10] DONE: got started with urbanape, adjusted some tests in the read_link issue to work with new NTFS hardlink impl [15:10] TODO: continue working with urbanape over ssh/screen/mumble on getting the buildout to work, receive my OS X install CD from FedEx and...install it [15:10] BLOCKED: none [15:10] NEXT: not dobey [15:10] λ DONE: rb nightlies fixes, bug #934235, [15:10] λ TODO: rb extension bugs, look at twisted g/gtkapplication support/hanging issue [15:10] λ BLCK: none. [15:10] alecu [15:10] DONE: cleanups on disconnections for bug #929207, debugging, fixing tests, natholiday on monday [15:10] TODO: rework bug descriptions for FFE, finish and submit branches [15:10] BLOCKED: no [15:10] NEXT: urbanape [15:10] DONE: Compiled PyQt ALL DAY LONG. [15:10] TODO: Hacking on the buildout with briancurtin [15:10] BLOCK: Nothing, so far [15:11] end? [15:11] any comments anyone? [15:11] urbanape: compiling Qt is a rite of passage. Wait for PyQt ;) [15:12] alecu, how are you writing those tests? [15:12] if not, I guess is eom! [15:12] ralsina: did PyQt [15:12] mandel, let me summarize what these tests are testing first: [15:12] DOUBLE RITE [15:12] urbanape: brave man! [15:13] mandel, SD needs to connect to the sync servers going thru the proxy tunnel. The tunnel listens for a few lines telling it where to connect, and then it forwards all traffic transparently. [15:15] mandel, but SD needs a "client" to connect to the tunnel and send the connection command (in plaintext) and then it needs that client to switch from the tunnel protocol to the ssl used to transport the SD traffic [15:16] mandel, so my tests are testing that transition: from plaintext (used to speak to the tunnel server) to SD protocol inside a SSL encrypted channel. [15:17] mandel, that's why I need to run a fake SSL server. And that SSL server needs a dummy certificate file and a dummy key file, that I generated manually and I'll be including in our tree. [15:17] alecu, ok, so that is a diff ssl scenario, I wanted to write tests for when the proxy that we are listening to is in a https port. I wanted to ensure that when we go over https and have a cert that is no vlaid we show the dialogs etc.. [15:18] wait, so we are going with an external proxy-handling process? Why don't we use some existing one? [15:18] alecu, I can provide you with code that will generate a random cert instead, is very easy with open ssl [15:19] mandel, cool, but I'm not sure if we'll have all the ssl cmdline tools in other platforms. [15:19] ralsina: can I have a review please? https://code.launchpad.net/~nataliabidart/ubuntuone-control-panel/work-on-windows/+merge/94980 [15:19] alecu, we have the library :) [15:19] rye, what existing proxy handling process do you propose? [15:20] alecu, http://paste.ubuntu.com/860638/ [15:20] alecu, some imports are missing, but they are from os, so its easy :) [15:21] nessita: sure, in about 10' [15:21] rye, the idea to build one was so that it uses the same library (QtNetwork) as the rest of our tools, so we have the same proxy support everywhere. [15:21] mandel, oh, that's lovely. [15:21] alecu, ah [15:21] https://wiki.edubuntu.org/RomanYepishev/UbuntuOne/ProxySupport :) [15:22] why is edubuntu wiki ranked higher in google searches than wiki.ubuntu.com ? [15:22] rye, because they are educated.. hehehehehe [15:26] rye: also, the last 2 points on that page are quite valid. and i suspect those tools don't work so well on windows? [15:33] brb [15:36] rye, proxytunnel is interesting. I evaluated using similar tools like cntlm, but each tool supported a different subset of options (ie, ntlm authentication, or socks proxies, etc) and supported different platforms, so we would end up depending on a handful of tools. [15:39] mandel, re: your generate_self_signed_cert() function; is it already on our sourcetree? [15:40] alecu, no, is work in progress of ubuntuone-dev-tools, feel free to use it and I'll deal with cleaning the duplication :) [15:41] mandel, great, thanks. [15:42] mandel, I think for now I'll use the certificates made by hand, but when this landsw on u1-dev-tools I'll switch to it, because my certificate will expire in 10 years :-) [15:42] alecu, hehe [15:43] alecu, I'll let you know when it does [15:44] nessita: +1 on work-on-windows [15:45] nessita: did you file a bug, or know of one, about the messaging menu opening the gtk control panel? [15:48] dobey, I saw it filed under bug #735464 [15:49] alecu: that is a separate bug, though people are commenting on it because they see similar behavior in 12.04 now [15:49] alecu: but it was originally filed on 11.04, and i suspect a different issue [15:50] right [15:50] popey, ping [15:50] popey, I'm trying to understand the bug with proxy hostnames: https://bugs.launchpad.net/ubuntu/+source/ubuntuone-client/+bug/942312 [15:51] popey, did you have proxies enabled on the system settings while this happened? [15:53] alecu: ooh, i think i see the problem. the dialog where you put your proxy.. if you have it set to "manual" you can leave the other fields blank and it still lets you save that [15:53] alecu: which IMO it shouldn't let you do [15:54] popey, well, each app could tell if the "hostname" was empty, and ignore the settings [15:54] popey, in this case it looks like libsoup is the bit where it's failing. [15:54] that shouldn't be necessary really === yofel_ is now known as yofel [15:55] I'll file a bug against the networky bit [15:55] also, unrelated to that [15:55] in rb I keep getting a prompt that "Mp3 playback support is not available" [15:55] dobey: the one about having -gtk mixed with -qt? I did not file it, nor I recall one about it [15:55] i click install and it goes off and does stuff but i still cant play music [15:55] (from the store) [15:55] and when i restart it says the same thing [15:55] is this a known issue? [15:56] nessita: yeah, the messaging indicator opening the gtk one when you click on it, and using the dbus interface to open the volumes tab [15:56] dobey: no bug in my radar for that [15:56] popey, no idea about that rb bug. Let's ask dobey. [15:58] alecu: no idea. we don't do any magic stuff for proxies there. if the same error comes up there, it could be coming from sso [15:59] dobey, no: I meant the "Mp3 playback support is not available" bug. [15:59] nessita: right, i have a fix; though it doesn't open to the "folders" section for the folders, as it just runs the installer instead (which in turn runs the qt control panel if installed) [16:00] dobey, the issue with proxies is when the user sets the proxies to "Manual", but leaves the proxy host field empty. [16:00] nessita: but i was looking for a bug to --fixes= for [16:00] alecu: ok [16:00] alecu: the mp3 bug is known/filed. i have to debug it still though [16:00] popey, ^ [16:00] ok [16:01] whilst I'm in rb i also have duplicates of all my u1ms music [16:01] i mentioned this previously and someone asked if i could get some info from my machine [16:01] but I dont know what I need to get [16:01] popey: yes, that bug is filed [16:01] ralsina: your opt-branch approved! [16:01] dobey, nessita: the u1sdtool bug is probably gi/twisted: it blocks waiting for data in a pipe (apparently) [16:01] nessita: cool, thanks [16:01] i know what it is. will probably fix it today, and fix will appear in ubuntu next week [16:01] ralsina: yeap [16:02] ralsina: nope. i have similar issues with other non-twisted things doing it as well [16:02] nessita: if you have a small bug or two, I may take a look while I do other stuff [16:02] dobey: I read what ralsina said as 'gi or twisted' [16:02] yes, one or the other :-) [16:03] ralsina: let me look [16:03] although it locks inside twisted, it could be caused by the interaction with gi [16:03] nessita: not sure. control panel qt has same issue [16:03] ralsina: we have this same behavior on other apps not using twisted [16:03] nessita: interesting [16:06] i think the qt control panel issue is probably easy to fix though [16:16] alecu, ralsina: for bugs #929207 #929208 #929212 we need to answer pitti's concerns (listed on 929207 & probably consolidate these into one u1 proxy support bug that we land the necessary branches against to make life easier for everyone [16:18] joshuahoover: got it, looking [16:18] joshuahoover: also bug #933729, no? [16:18] ralsina, mandel: ^ [16:19] nessita: yes! you beat me to my next one :) [16:19] :-) [16:19] mandel, ralsina: bug #933729 (http://pad.lv/933729 ) has a comment from pitti w/ a concern he has about handling ssl mismatches that we need to answer [16:20] joshuahoover, alecu, nessita: I guess the answers are "no", "because syncdaemon can't depend on (what?)" and "the system, except for authentication which is not standard"? [16:20] joshuahoover, well.. his concerns are right, we are going to let the user be stupid [16:20] ralsina: regarding the tunnel process, I know alecu will answer during the day [16:20] nessita: cool [16:20] ralsina, right? [16:21] joshuahoover, I agree about consolidating some of the proxy bugs [16:21] nessita, ralsina: those are the FEs i found that needed some additional info/help...the others look like they'll be ok assuming we get them landed soon...i need to follow up on one other http://pad.lv/933010 which is a uife that i didn't notify the doc and translation lists about yet [16:21] The only possible source for proxy ssl mismatches is that the user misconfigures the proxy name [16:22] ralsina, or the it people that set up the proxy certs are stupid, I have worked in one of those.. [16:22] But for example, you will get a mismatch if you configure the proxy's IP address instead of the canonical name, so there *may* be benign mismatches :-( [16:25] mandel, alecu, joshuahoover: So, basically, I don't care as long as we give a good explanation about the mismatch. Supporting misconfigured setups is not a priority. [16:25] ralsina: what id you tackle this on you reported? bug #942355 [16:25] what if* [16:25] (ralsina: speaking of what you asked beore about doing some bugs fixes on the side) [16:25] nessita: ok [16:25] nessita: should not be terribly hard :-) [16:26] ralsina: I'll assign that one and another one, JIC [16:26] nessita: cool, thx [16:26] ralsina: though... this one requires a UIFe https://bugs.launchpad.net/ubuntuone-control-panel/+bug/942025 :-/ [16:26] joshuahoover: we're only supposed to notify docs/translations (probably only translations for this one), after approved though, right? [16:26] so, no, that one no [16:27] nessita: I can do the branch then we decide if we ask for it or not [16:27] dobey: we have to do it before approval for all UIFEs...to let them know about it and allow them to express their concerns [16:27] ok, not that one :-) [16:27] ralsina: I'm talking about another bug (not the one about the scrolling) [16:27] nessita: the email address font size one? [16:27] yes [16:28] so, the scrolling is bug fix [16:28] the font-too-small is UIFe [16:28] ack [16:28] ralsina: I assigned the scroll-fix to you [16:29] nessita, ralsina: i think http://pad.lv/942025 for the small font size is ok to fix w/o a ui exception but that's just my opinion and definitely not following all the rules ;) [16:29] joshuahoover: hum, how do you justify it does not require a UIFe? [16:29] joshuahoover: is changing a "screenshot" of that tab [16:29] small? [16:29] dobey: right, "small" :) [16:29] all the fonts in that app are freakin huge [16:29] trivial and obviously wrong and of no consequence? ;-) [16:30] and, for some reason, bold. [16:30] oh well, i should get some lunch [16:30]