=== eu is now known as Guest45697 === eu is now known as Guest81574 [08:01] morning all! [08:25] Good morning all! :) [09:44] dobey, I found the issue, the problems is that QNetwork.setApplicationProxy does not work and we have to set the proxy on the NetworkAccessManager, which will mean that the code won't work in older versions.. [09:44] dobey, lets talk about the best approach whenever you are around === mpt_ is now known as mpt === eu is now known as Guest32810 [11:21] good morning [11:23] good morning here too [11:32] gatox, morning to both of you [11:32] mandel, jeje thx [11:32] mandel, i'm in Q now.... and it's really fast [11:32] successful upgrade [11:33] gatox, yet proxy support for u1 is broken due to an update of qt [11:33] :P [11:33] ouch [11:35] gatox, might be a big deal if we have to be compatible with older versions [11:35] gatox, they fixed a bug that broke a workaround hehe [11:36] mandel, jejeje well... we might have something as we do for the multiplatform code..... based on the qt version..... is a PITA but it will work [11:36] gatox, which will be a fuuuuu from my point of view ;) [11:36] yap [11:37] * gatox is really happy about how Q is working in his machine :D === eu is now known as Guest80106 [12:28] fuuuuuu http://paste.ubuntu.com/1257851/ [12:43] good morning! [12:43] ralsina, hi! [12:44] ralsina, 12.10 is working like a charm...... just to let you know [12:44] gatox: awesome! [12:44] ralsina, morning! [12:44] hello gatox, mandel [12:45] ralsina, I hate unity => http://paste.ubuntu.com/1257851/ [12:45] ralsina, also, QNetwork.setApplicationProxy() is not working in the new qt as it did in the old one, it is easy to fix, but.. which qt version are we using in old ubuntu distros? [12:46] mandel: 4.8.something [12:46] mandel: let me check [12:46] ralsina, will it get updated, do we have to behave in a diff manner according to the qt version? [12:46] mandel: I am guessing it will get updated in 12.04.1 [12:46] mandel: but who knows. In any case, we should still try to support the old way if it's not a huge mess [12:46] mandel: 4.8.1 [12:46] ralsina, ok, I'll talk with dobey on what to do, is as simple as not using the application proxy and set the proxy for the networkaccessmanager [12:47] ralsina, we can always have an ugly if statement [12:47] mandel: you are missing a file in a build dep, obviously [12:47] mandel: exactly [12:47] mandel: if it's just *one* if, I don't mind [12:48] much === Guest80106 is now known as adorilson [12:49] ralsina, should not be.. but might have to check for version and os.. :( [12:50] ralsina, so is not as nice [12:50] mandel, do you were involved in the code for SyncDaemonTool signals handling? [12:50] gatox, nop, I might know something but I was about to change the nick to mandel|lunch :P [12:51] mandel: hmmmm version & os? Really? [12:51] mandel, no problem..... it wasn't to ask you something now...... i'm still trying to figure it out something..... i just wanted to know who to ask if this get more awful :P [12:51] ralsina, we need to check [12:51] ralsina, I have not tested it on windows [12:52] gatox, I think I can give you a hand [12:52] well, I'm off to have lunch === mandel is now known as mandel|lunch [12:55] mandel|lunch: hopefully, we try it on windows and it's broken in a cross-platform-compatible manner [12:59] mandel|lunch: hrmm, can we just make all the calls all time, so it works on both versions regardless? [12:59] mandel|lunch: is there documentation on why this broke? this seems like a regression in qt, so perhaps we can just push a fix to it? [13:00] dobey: good idea, should not break anything setting it twice === yofel_ is now known as yofel [13:09] hello #ubuntuone! [13:09] hola alecu [13:11] Hello, fellow desktoplings! Vengo en paz. And with a MP: https://code.launchpad.net/~facundo/ubuntuone-client/fix-path-retrieval/+merge/127583 (thanks!) [13:11] hey all, I'm taking the morning off to pay another visit to the doctors [13:11] * alecu aims his raygun to facundobatista [13:12] alecu, :) [13:16] facundobatista: ya casi terminé el review de esa [13:18] ralsina, maravilloso, gracias [13:20] facundobatista: +1 [13:20] ralsina, :D [13:20] ralsina, thanks [13:26] I am rebooting to windows to help test some u1db stuff, will be back in a bit [13:41] Damned, I had forgotten how painful it is to boot windows if you have not done it in a while [13:43] gatox: since you just upgraded to 12.10, did your u1 icon in launcher go away? [13:43] ralsina, i didn't have it there anyway..... so i don't know [13:44] gatox: could you create a fresh accout and see if it's there? [13:44] ack [13:44] brrb [13:44] gatox: just switch to the guest account :-) [13:45] ralsina, the guest session is enough? [13:45] yes..... it's there === mandel|lunch is now known as mandel [13:56] ralsina, dobey well, is not that easy, we used the setApplicationProxy over the NetworManager.setProxy because in previous versions of qt there was a bug in it (setProxy) [13:56] ralsina, dobey you could not set the proxy with the correct auth (after the first requests which has no auth) because it would be cached in the nam [13:57] we would need to test if setting both is the correct approach, yet i'm 100% sure that if you set it in the nam the tests will pass in Q [13:59] gatox: can you doube-check that it's the control panel and not the music store? Just in case :-) [13:59] ralsina, i have both [14:00] ralsina: i have seen it go away on upgrade though [14:01] dobey: right, that is what I got reported. It seems to go away for accounts that already existed? [14:02] ralsina: it goes away if you upgrade while logged in, i guess [14:02] dobey: makes a little sense [14:02] dobey: any idea who we should ping about it? [14:03] ralsina: not sure [14:03] ralsina: i guess dx probably, as it might be a bug in bamf [14:04] dobey: I am doing other stuff right now and in windows to boot, could you report the bug? [14:05] going to test upgrade on other laptop first [14:05] dobey: awesome [14:12] i hope upgrading doesn't break the world on that machine though. (intel gma500) [14:22] hrmm, i think maybe the issue is that ubuntuone-installer gets removed at the beginning of the upgrade, and and new control panel isn't isntalled until much later (which could be an hour apart or something), and so as far as unity is concerned, the thing was in fact removed :-/ [14:23] but it would only in fact affect users who run the upgrade while logged in as their user [14:23] so you know, everyone. [14:25] and the "base-files" just failed to install for me. whee :-/ [14:27] dobey, i can try to upgrade my other laptop if you need [14:29] well it's obvious what happened [14:29] i am just not sure it's possible to fix it [14:33] dobey: well, then we have to tell DX I suppose and see if they figure it out [14:41] dobey: I am asking in ubuntu-desktop let's see what happens [14:44] i can't even find where that list of launcher items is even stored, any more === m_conley_away is now known as m_conley [14:55] gatox: did you get to look at those control panel tests failing? [14:57] dobey, hop… i'm kind of stuck with the u1-cp issue in mac yet [14:57] going back to linux, see you all in a bit or two [14:58] gatox_mac: can we postpone that to fix the tests issues? [14:58] dobey, fine by me if ralsina agrees [14:59] ralsina: ^^ [15:00] me [15:00] me [15:03] ah crap, typing now [15:04] me [15:05] me [15:07] me [15:08] mandel, go [15:08] DONE: FOund out what changed in qt and broke proxy. QNetwork.setApplicationProxy does not work as it used to. We need to set the NetworkAccessManager proxy instead. Merge u1 payment preview with tunk, everything went down to hell [15:08] TODO: Try to fix the merge problems which stops my code fomr compiling :*( Propose fix for proxy. [15:08] BLOCKED: no [15:08] gatox, please [15:08] DONE: [15:08] Fight a lot with twisted and u1-cp signals for the public files list. [15:08] TODO: [15:08] Figure it out why the last part of the signal is not reaching the share_links module, although we are getting the correct data from u1-client. Fix u1-cp tests [15:08] BLOCKED: [15:08] Trying to understand some part of the twisted signaling implementation. [15:08] mmcc, go [15:09] DONE: tests for cmd-q catch, daemon crash debugging [15:09] TODO: merge cmd-q, daemon work [15:09] BLOCK: no [15:09] NEXT: briancurtin [15:09] DONE: well U1CP starts up now, usually, in most cases, somehow. [15:09] TODO: figure out what's still in the way of it operating, possibly the recent changes to finding the binaries [15:09] NEXT: dobey [15:09] DONE: releases, uploads [15:09] TODO: finish 4.0.0 releases/uploads, poke py3 twisted guys about packaging [15:09] BLCK: New Qt broke SSO authenticated proxy support. CP tests breakage. [15:09] ralsina: go [15:14] grmbl, sorry, ubuntu update broke network :-( [15:14] ralsina, yep..... here too..... but it get connected eventually (specially if i'm close to the router.. magic :P) [15:15] dobey, I'll propose a fix for that block asap, let me finish first with some work on unity [15:15] dobey, I'll try to test it in P and Q [15:15] ralsina, so..... about what dobey says...... [15:15] ralsina, gatox_mac: can we postpone that to fix the tests issues? [15:15] DONE: calls, helped u1db test on windows, some reviews, helped around a bit TODO: try to figure out that upgrade problem, help around some more, etc. BLOCKED: no [15:15] gatox: how stumped are you on the mac u1cp stuff? [15:18] ralsina, the problem seems that we are getting all the info from u1-client..... i'm printing that and i see it..... but in some place in the endless callbacks to return the signal from u1-client, one of the callback is not getting called and the method which receive the data and process the info in share_links is never being called...... and i'm not really familiar with that part of the signal and wait_for_signals stuff in u1-client and i do [15:18] n't quite understand yet what is happening with a deferred that set a callback to i-dont-know-what [15:19] gatox: ok, you may get some help from mandel for that tomorrow morning [15:19] gatox: so, switch to the tests now, and then you go back to the shares tab tomorrow early [15:19] ralsina, i was about to execute the same thing on windows.... that uses the same calls...... and try to follow everything backwards with the stack..... to see what funcions should be called on the response [15:19] hm.. ralsina may or will? :-) [15:20] mandel: I am not the "thou shall" kind of manager, so add a "?" there where appropiate ;-) [15:20] hahahahaha [15:20] ralsina, I suspected that :) [15:20] mandel: I really hope you will though [15:21] will do my best [15:21] ralsina, ok..... i'll move to the tests.... then try the windows option to see if i can track which should be the proper path of calls...... and if i start to have suicidal thoughts..... i will ask mandel :P jeje [15:22] gatox, for how to kill yourself? I would go for alcohol and unprotected sex [15:22] jejejeje [15:24] hey guys, I need to do some emergency babysitting now, sorry. I'll peek in occasionally but might be gone for an hour [15:24] (my wife is sick this morning) [15:24] mmcc: sure, go ahead [15:25] gatox: if you are going to be testing things in windows from source, make sure you already have creds otherwise you wont be able to start U1CP [15:27] briancurtin, ack, thx...... and i have creds [15:28] ok, i need to get some lunch. bbiab [15:32] * gatox lunch....... brb === gatox is now known as gatox_lunch [16:02] lunch for me too === gatox_lunch is now known as gatox [16:27] lunch + quick_errands = brb [16:28] EOD for manuel [16:28] aka me [16:32] mandel, bye [16:35] so just need to get sso and control-panel released and uploaded, but grrr qt [16:53] dobey, where are the tests failing? llinux? [16:54] gatox: yes, in taramc [16:54] gatox: https://code.launchpad.net/~dobey/ubuntuone-control-panel/update-4-0/+merge/127559 [16:54] dobey, ahhhhhh it was client...... i was confuse [17:15] mandel: you had a branch? [17:18] dobey, do you know if there is a bug for the failing test or it should be? [17:18] gatox: i haven't filed one yet [17:18] dobey, ok.... i'll do it and propose the branch now [17:19] gatox: ok, thanks. what was the problem there? [17:20] dobey, some part of the test wasn't updated with one of the latest changes adding the separator, which counts as another item [17:21] gatox: a branch isn't needed i guess [17:21] if that's the issue [17:21] uh, wait, what does that have to do with control-panel? [17:21] dobey, do you want to add the patch in your branch [17:21] dobey, no, i'm talking about the failing tests in u1-client [17:22] what failing tests in u1-client? [17:22] dobey, there was a test failing in u1-client.... i thought you were talking about that one [17:22] gatox: no i'm talking about control-panel [17:23] dobey, the one in u1-cp is failing only in mac..... at least for me [17:23] gatox: a lot of tests are failing with regards to the shares tab [17:23] i'll double check right now [17:23] gatox: look at the failure message here: https://code.launchpad.net/~dobey/ubuntuone-control-panel/update-4-0/+merge/127559 [17:24] dobey, yes.... i saw that..... but i get all green on linux...... and that only on mac [17:25] that's why i was working on mac [17:25] gatox: on precise, or quantal? [17:26] dobey, now i'm on quantal..... but yesterday i was in precise and i had u1-cp tests green..... that's when i notice the tests failing on mac and not on linux [17:26] and i said the part about being consistent.... [17:26] because for some reason that is not working on mac [17:27] gatox: they pass for me on precise too, but these failures are on quantal in tarmac [17:28] dobey, weird.... [17:28] gatox@utopia:~$ lsb_release -a [17:28] No LSB modules are available. [17:28] Distributor ID: Ubuntu [17:28] Description: Ubuntu quantal (development branch) [17:28] Release: 12.10 [17:28] Codename: quantal [17:29] * briancurtin back [17:31] dobey, if you can...... this fix the failing test on u1-client: https://code.launchpad.net/~diegosarmentero/ubuntuone-client/ubuntuone-client-fix-tests/+merge/127836 [17:31] dobey, now i'll go back to mac..... there i'll be able to fix those broken tests..... plus that tab not working there [17:34] gatox: ok; you are getting those same errors on mac? [17:34] dobey, yes..... only there [17:35] hrmm [17:41] mmcc: ping [17:41] just got back. what's up chaselivingston ? [17:42] mmcc: just created this http://bazaar.launchpad.net/~chaselivingston/+junk/scripts/view/head:/mac-logs.sh [17:42] what's the best way to distribute that for people to use relatively easily [17:43] you can make a double-clickable command file that opens in the Terminal. letme look and see how [17:43] ok cool, that would be helpful [17:44] also, if you could add /Library/Logs/DiagnosticReports/com.ubuntu.* to that zip that'd be good too :) [17:44] that's where crash reports for the daemon are saved [17:45] can I add that to the same line somehow? [17:46] yeah, if you just add it to the end it should work [17:46] ok, i'll give that a shot [17:46] ok, so if you just make that .sh file executable then people can double-click on it. I thought you had to use the .command suffix, but the help says you don't [17:47] mmcc: so how do i make it executable [17:47] chmod +x mac-logs.sh [17:47] ok cool [17:48] btw, you might not have any crash reports in /Library/Logs/DiagnosticReports for the daemon (actually I'd like tos ee them if you do) [17:48] that file path you sent for the daemon logs isn't working, not sure how to grab individual files, although i do see them in there [17:50] and running chmod +x mac-logs.sh doesn't seem to do anything... [17:50] hrm [17:50] so, 'ls -l /Library/Logs/DiagnosticReports/com.ubuntu.* ' shows files ? [17:50] yes [17:51] chmod is quiet if it worked, btw. check ls -l mac-logs.sh to see if there's an 'x' in the permissions string [17:51] like this: -rwxr-xr-x 1 mmccrack staff 8.7K Aug 29 13:36 setup.py (the permissions are at the beginning) [17:51] yes, there is [17:51] ok, so it's executable. try double-clicking it [17:51] it just opens in xcode [17:52] er, ugh [17:52] (or dashcode, maybe) [17:52] what happens if you rename it mac-logs.command [17:52] and doubleclick that [17:52] one sec... [17:53] that works [17:54] mmcc: now i just need to figure out how to grab those other crash reports [17:56] wtf, there are 7 of the not so great orange u1 windows icon on my panel now, while running the tests on quantal [17:56] mmcc: i can get the entire DiagnosticReports directory, but not those individual files [17:57] chaselivingston: what command are you using, verbatim? This just worked for me: zip -r ~/Desktop/logs.zip ~/Library/Caches/ubuntuone/log/ /Library/Logs/DiagnosticReports/com.ubuntu.one.fsevents* [17:57] thankfully they've all gone away [17:58] hrmm [17:58] mmcc: hm, worked that time w/ what you sent [17:58] don't think i had fsevents on there previously [17:58] whee, and pylint is leaking like crazy [17:59] mmcc: just pushed the latest revision [17:59] 400M RES so far [17:59] chaselivingston: ok… com.ubuntu.one* will match all the files that com.ubuntu.one.fsevents* does, but as long as it works [17:59] i'm surprised pylint doesn't get OOMed on tarmac [18:00] mmcc: yeah, well there we go. we'll probably add an faq about that once the beta goes public [18:00] that way people can easily attach logs to bug reports [18:01] chaselivingston: yeah, helpful. thanks! [18:01] mmcc: np [18:01] chaselivingston: btw, you might also want to know about the shortcut "Control-Option-Command-Shift-Period". it runs 'sysdiagnose' - https://developer.apple.com/library/mac/#documentation/Darwin/Reference/ManPages/man1/sysdiagnose.1.html [18:02] mmcc: interesting, i'll take a look, thanks [18:02] that's kind of a last resort, but if you suspect something's weird with their system, that gives you all kinds of info, including apps installed, OS version, running apps, etc etc [18:02] ok great [18:03] kind of intrusive though, so not usually something I'd ask for. just worth knowing about [18:04] mmcc: right [18:13] weird, the tests passed for me on quantal :( [18:26] mmcc: fyi, http://people.canonical.com/~chaselivingston/scripts/mac-logs.dmg [18:26] chaselivingston: thanks [18:30] Here's one for a blog post: unicode test path in our control-panel tests was crashing the fsevents daemon! [18:31] mmcc: is there a command to quit terminal after that process has run? [18:31] chaselivingston: I don't think so. You wouldn't want it to quit if there were other windows open, for instance [18:31] mmcc: uh? [18:32] mmcc: ok gotcha [18:32] mmcc: Oh, I get it! AMAZING :-) [18:32] ralsina: it's great! I was getting a lot of daemon crashes and wondering why no one else was… [18:32] mmcc: that's becaue you are the only one running the crazy filename tests with the daemon running :-) [18:33] was also wondering why I got a ton of crashes on monday afternoon and none over the weekend… :) I even left Spotify on over night thinking that might be it [18:33] mmcc: so yes, that's totally blog-worthy [18:33] added a bunch of debug dumps and figured it out just now when I started running the tests again. Hooray for having a clue! [18:47] another funny twist, the 'mdworker' system process that does indexing for spotlight also complains about the path that was crashing our daemon… [18:50] ralsina, could you review this one: https://code.launchpad.net/~diegosarmentero/ubuntuone-client/ubuntuone-client-fix-tests/+merge/127836 ? [18:50] gatox: por supus [18:50] ralsina, thx [18:51] mmcc: hey, a corner case! ;-) [18:56] ralsina: well, this is interesting. it's in test_folders.py where we add '\xe7\xa7' to a file name before opening it and writing to it… I'm not clear on what that's testing, and if it's something that'll actually happen [18:57] ie, I can just ignore the event if it has an invalid filename, but we shouldn't if that happens in real life [18:57] mmcc: I *think* we are testing invalid unicode [18:57] mmcc: which yes, should not happen IRL but does because people have ancient files from dubius zip files and such [18:58] btw, I recreated that test and crashed the daemon again, but it did create a file - it shows in the finder as %XE7%XA7 [18:59] the place where it causes problems in the daemon, it's expecting a utf-8 filepath from the fsevents device and it can't translate it [19:00] hmm [19:00] mmcc: what test is this? [19:00] mmcc: yes, that's a "invali utf8 filename" [19:00] mmcc: so it's a real bug in the daemon [19:00] dobey: class CalculateSizeWithInvalidPath(CalculateSizeTestCase): in test_folders.py in controlpanel [19:00] mmcc: since it seems fsevents sometimes passes things that are *not* utf-8 paths [19:01] mmcc: what's the full string for the path that's failing? [19:01] yeah, it's a real bug for sure. [19:02] dobey: if you open("\xe7\xa7", 'wb') from the python console, it'll poke this . but it's not failing the test, it's causing a failure in the daemon [19:02] the tests are fine :) [19:03] Hey guys, does our Windows client need to replace some file system unsafe characters when downloading files form U1? (Say, someone has FAT or VFAT partition) [19:03] karni: yes [19:03] Or do we assume any sensible person has NTFS? [19:03] karni: lots and lots of them [19:03] ralsina: oh, tell me more! [19:03] karni: on windows, even on NTFS there are invalid characters [19:04] karni: that includes * ? : and others [19:04] ralsina: http://en.wikipedia.org/wiki/Filename says / and null only o_O [19:04] karni: mandel is the expert in the fiel [19:04] d [19:04] mandel: still around? [19:04] karni: invalid for the filesystem and invalid for the OS are different things :-) [19:05] ah "forbids the use of characters in range 1-31 (0x01-0x1F) and characters " * : < > ? \ / | unless the name is flagged as being in the Posix namespace. " [19:05] karni: for example, you can't create a file called AUX.SVG IIRC [19:05] right [19:05] I've read that up, yes. [19:05] ralsina: Maybe you'll be able to answer the following question for me: [19:06] karni: I will try! [19:06] if we download a file, say we replace those unsafe characters with _. we may cache the resource path it came from, but what if the user reinstalls U1. do we upload those files back to U1 with underscores? [19:06] karni: yes because we have no way to know [19:06] so that foo* and (reuploaded foo=)foo_ will appear along in U1? [19:06] fuck.. [19:07] ralsina: That answers my question, thank you! [19:07] karni: if we tried to be that smart, then we fail in the opposite way [19:07] ralsina: That's what I thought :( [19:07] mmcc: was that the full path? i'm just asking because the \xe7\xa7 seems a bit familiar [19:07] karni: basically, we are losing information, we can't just create it when it moves the other way [19:07] ralsina: Exactly. As simple (and sad) as that. [19:07] karni: same thing about upper/lower case :-( [19:08] ralsina: Yeah :/ what do we do with that on VAT? Say, user as foo and FOO in U1. What syncs down? [19:08] dobey, no the full path was /my-path/to/_trial_temp/not-existing-dir/\xe7\xa7 [19:08] karni: sorry for foul language, but basically, shit happens [19:08] ralsina: omg :< [19:08] karni: we got a report of foo vs Foo today [19:08] FAT, go away! [19:08] karni: the user ended with both on ubuntu [19:08] and only one on windows [19:08] ralsina: Right... [19:09] I am scaredd to ask what happened to the files inside the folder [19:09] ralsina: Did he/she say what they expected (knowing the limitation) ? [19:09] karni: AFAIK "magic" [19:09] besides having foo and FOO appear on Windows (which they can't) [19:09] :/ [19:09] karni: I mentioned the problem with case sensitivity, no response yet [19:09] ralsina: I feel like you feel my pain. Thank you for sharing that, it's easier for me that way ;< [19:10] ah ok [19:10] karni: np [19:10] karni: buy mandel a beer in the sprint, he has stories [19:10] ralsina: btw. if foo and FOO are in U1, and, say, foo gets downloaded, then FOO gets downloaded (overwriting locally foo), foo (having FOO content) will overwrite foo when reuploaded to U1 o_O?! [19:10] ralsina: totally [19:11] karni: probably [19:11] oh shit [19:11] karni: or maybe it will get even more confused and delete everything. Dunno. [19:11] karni: wanna try? ;-) [19:11] ralsina: hahaha [19:11] ralsina: I.. I don't think I have windows :| [19:12] ralsina: Anyways, thank you for confirming these issues. [19:12] karni: chicken! ;-) [19:12] LOL :D [19:12] karni: as a punishment https://answers.launchpad.net/ubuntu/+source/ubuntuone-client/+question/210270 [19:13] you need to rename it "u1-files-which-doesn't-sync-your-changes-to-the-server.apk" [19:14] ralsina: f!@# I WILL! [19:14] chaselivingston: ↑ Shall I write a FAQ for you to add to our FAQ's page? [19:14] ralsina: FWIW there's a community person working with me to get things spinning around real sync in U1F. [19:14] karni: yeah, you could do that, not a bad idea [19:15] ah windows [19:15] ralsina: But GOD we never EVER said it does SYNC. [19:15] I should review Android market description, this smells. [19:15] karni: but it's files! they should obviously sync! [19:16] dobey: it's an expectation from desktop U1, yes [19:16] dobey: FWIW I wanted to implement it long ago. There are just priorities I have to follow. [19:16] karni: does dropbox on android automatically push changes back to the server too? or box.net? or any of the others? [19:17] karni: sorry if you couldn't detect the sarcasm in my voice over text. clearly some information was lost. ;) [19:17] dobey: only if you open that file for edition from within the app itself, which I found plain stupid (read, limiting) [19:17] dobey: :D [19:17] dobey, karni; would be pretty cool if we were the first to implement syncing like this well [19:17] dobey: I never wanted to deliver sync solution to U1F that is half baked. I would hate that. [19:18] chaselivingston: We have some work done already, we're in good direction. [19:18] karni: yes, i agree [19:18] karni: good to know [19:18] if a file appears or is changed, in the background, not even touching U1F, U1F should still detect it. [19:18] but i also don't use android :) [19:18] Anyways, gotta talk to Chad about (V)FAT. [19:18] dobey: hehe [19:23] So I have a problem - if I can't represent a file path in utf-8 then I can't really send it to syncdaemon in JSON… [19:24] mmcc: right; but you can log/ignore it [19:25] dobey: which means that the path won't get synced. How does syncdaemon handle invalid paths in its scan code? maybe I can send a fake dropped event [19:27] mmcc: yes you can, json can encoe arbitrary bytes [19:28] ralsina: well, it'd take a bunch of reworking of the json writing code in this library we're using… [19:28] it just blindly writes everything out as utf8 [19:28] mmcc: we don't sync them afaik [19:28] mmcc: I take that back, json can't [19:28] sheesh [19:29] everything breaks [19:29] mmcc: I suppose we could just ignore all events regarding invalid filenames [19:29] mmcc: they are, by definition, a broken corner case [19:29] ralsina: I think that's the way to go for now, yes [19:30] in the future we could base64-encode the bytes paths (ugh) [19:30] btw, dropbox silently ignores a file with invalid filenames... [19:30] er, with an invalid name [19:30]