[00:07] mwhudson: What version of pygments is installed on codebrowse? [00:10] mkanat: whatever's in hardy i guess [00:11] mwhudson: Okay. [00:11] mkanat: 0.9-2 [00:11] mwhudson: Thanks! :-) [00:11] * igc building Windows installers for 2.2b2 today [00:38] krisives-gearbox: well, the old commit would still be there.. you're creating only a new one. I would go with several "uncommits", fix the one in question, then redo the latter ones. edgimar ^^ [00:40] spiv, bug 496813 has landed into 2.1 now right? [00:40] Launchpad bug 496813 in bzr/2.1 "until_no_eintr used on unrestartable IO, and cannot address all cases of EINTR." [High,In progress] https://launchpad.net/bugs/496813 [00:51] hi all [00:59] igc: yes, thanks, bug updated [01:00] spiv: thanks [01:04] lifeless (if any), I was thinking last night [01:04] perhaps we could test pqm in the real environment but on a different branch [01:04] so that it doesn't block development like bug 568100 [01:04] Launchpad bug 568100 in bzr "PQM broken for bzr/2.0" [High,Confirmed] https://launchpad.net/bugs/568100 [01:07] poolie: so thats not pqm itself; thats the subunit version in the chroot [01:08] because it changed recently and 2.0 is incompatible? [01:08] ok, but even beyond that specific bug [01:08] the lack of feedback is hopefully fixed at this point, or at least, one less cause removed [01:08] poolie: an edge/testing environment would be good [01:08] I'm not dismissing that [01:08] needing to test it in the deployed environment seems not to be quite the same as needing it to be tested within our main workflow [01:08] just noting the details of that particular problem [01:09] so we need to land a change to 2.0 that makes it pass with current subunit? [01:09] yes [01:09] i see [01:09] I've commented with a reasonable pointer in the bug [01:10] I think an edge/staging environment for our CI stuff would be great [01:10] that way when we move to tarmac, or whatever changes we want, we can be more confident before making changes. [01:12] so that would be i guess a different mail address and a different destination branch? [01:12] and different test chroot [01:12] and different pqm tree [01:13] possibly even a different machine [perhaps a UEC vm] as there are dependencies involved like launchpadlib version [01:33] hi poolie. Trip back to the big smoke all go ok? [01:34] it was fine [01:34] i really liked seeing brisbane again [01:34] Hi, does anyone have any pointers on integrating NTLM support? I've been using python-ntlm and urllib2 directly in my TFS plugin and would like to switch to using the HTTP transport. [01:34] and seeing you [01:34] poolie: we even put on some rain for you :-) [01:35] poolie: ditto [01:35] greetings everyone. [01:36] I was wondering if anyone would have some time to discuss the workflow for a standard bazaar development environment [01:36] lifeless: is it possible at all that pqm could put the failures on its own webserver and just post a link to the mp? [01:36] Mostly I want to make sure I interact correctly with the credentials support so I can prompt for a password; should I just look at the urllib2 wrapper stuff? [01:37] jbowtie: it sounds good; you probably want to come in through the bzr authentication layer [01:37] I have successfully used bazaar in both a local and bzr+ssh fashion, but when I use the bzr+ssh fashion it always commits to the server rather than locally. [01:37] but ntlm has an intimate connection to http so it may be complex [01:37] j: vila knows the most about this but he may be asleep [01:38] lifeless (or whoever): The lru_cache thread fix is in 2.1.0, right? [01:38] (That's what the bug says, I just want to be sure it's right.) [01:40] poolie: I was hoping I could just plug in the urllib2 provider that python-ntlm provides, but since all that stuff appears to be wrapped I was unsure how straightforward it was going to be. [01:41] rephormat1: I usually create a branch locally, commit there, then push to the remote server. [01:43] If we get NTLM support integrated I can publish my bzr-tfs plugin, that's really the only missing piece. [01:50] * igc out for a few hours [01:51] (2.2b2 builds for windows kicked off) [01:55] j: does that mean you would do a bzr checkout bzr+ssh://server/project/ [01:55] J: make changes to the code, test, commit [01:55] j: then do a bzr push? [01:59] rephormat1: Well, I would do a bzr branch; if I did a bzr checkout I wouldn't need to push (bound branches commit to both the server AND the local branch) [02:00] ahhh [02:02] jbowtie: so I would do a bzr branch bzr+ssh://server/path/project ~/project [02:02] jbowtie: make my changes, test, commit, then bzr push? [02:04] rephormat1: Yes - with that workflow everything stays locally until you push to the server. [02:04] mkanat: +1 to working on bug 420738 [02:04] Launchpad bug 420738 in loggerhead "LRUCache.cleanup raises KeyError" [Critical,Confirmed] https://launchpad.net/bugs/420738 [02:04] EXCELLENT! [02:05] jbowtie: What happens if I make changes to the same file as one of my other editors? [02:05] jbowtie: just tell me if I'm being lazy and I'll try to make sense of the documentation again. [02:05] mwhudson: Thanks. :-) [02:07] rephormat1: Usually that stuff just merges seamlessly. If there was a merge conflict I imagine you'd have to pull down the server changes, resolve the conflicts locally, then push again. [02:08] jbowtie: Thank you very much for your help, jbowtie. I appreciate it. Trying to get my programming group at our university to switch to bazaar [02:08] rephormat1: I have Subversion habits so I usually pull just before pushing as a safety precaution, but bzr has never bitten me. [02:09] Now all I have to do is find a bug tracking web gui for bazaar [02:09] rephormat1: No worries, doing the same at work which is why I'm working on a TFS plugin. [02:09] rephormat1: Bug-tracking and version-control are unrelated.... [02:09] I can't do it on launchpad.. at least not right now as all this code is inhouse and can't be public [02:09] TFS? [02:10] TFS = Team Foundation Server. Horrible Microsoft proprietary thing. [02:10] mkanat: my apologies. What I meant was being able to reference a bug ticket like you can when doing a commit with launchpad. [02:11] jbowtie: DUDE! I feel for you man. [02:11] rephormat1: Ahhh. bzr can do that with Bugzilla out of the box. [02:11] jbowtie: I finally got my group AWAY from SHAREPOINT!!!!!! [02:11] mkanat: YOU ARE AWESOME!!!! [02:12] Now if only I can kill off the microsoft group. LOL!! [02:12] rephormat1: Although I don't currently know of any integration within Bugzilla for bzr. [02:12] rephormat1: That might be my next project after this. [02:12] mkanat: I don't understand. Your saying it works out of th box, but that there is no integration? [02:13] rephormat1: bzr supports noting that a particular commit is associated with a particular bug in Bugzilla. [02:13] rephormat1: I think Trac is also well supported; there's a bzr plugin IIRC. [02:13] rephormat1: It's a feature in bzr. [02:13] Can you imagine.. a primarily open-systems group using SHAREPOINT!!! [02:13] so I showed them the light with Mediawiki [02:13] LMAO!! [02:14] jbowtie: yeah I was looking at bzr-trac at launchpad. The only thing I didn't like about it was that it required a directory to be completely empty to create a "trac"d branch [02:15] mkanat: I'm going to have to research that. Do you know if bugzilla can be downloaded and setup as a local private bug reporting system? [02:15] just curious.. [02:15] rephormat1: Yes, for sure. [02:16] mkanat: I TRIED to install a local private setup of launchpad... MAN that was a pain in the a$$.. [02:16] rephormat1: I can imagine. [02:17] mkanat: so I gave up... I'm a quitter.. I know... lol [02:18] but both of you have just peaked my curiousity and made my convincing arguement that much easier [02:18] Okay, I have spent six hours reading this effing log, it's time to take a break. [02:19] mkanat: What log? [02:19] rephormat1: Unrelated to our conversation. [02:19] rephormat1: Related to loggerhead work I'm doing. [02:19] mkanat: excellent [03:07] Hi [03:08] I need help [03:08] I was trying to download the Ubuntu Manual project files [03:08] via bzr branch lp:ubuntu-manual last night [03:09] I was able to start it, but in the middle of the download, I decided to cancel (since I'm going to bed already) [03:09] Now, when retrying to download it gives me "bzr: ERROR: Connection closed: Unexpected end of message. Please check connectivity and permissions, and report a bug if problems persist." [03:09] How do I correct this? [03:10] I've tried with different directory/folder (with the same name, and with a different name), but still the same error [03:11] zeroseven0183: So you're trying to branch it again? [03:11] Yes, correct [03:13] "Permission denied (publickey). bzr: ERROR: Connection closed: Unexpected end of message. Please check connectivity and permissions, and report a bug if problems persist." [03:13] zeroseven0183: OK, I seem to be branching it without a problem. Can I assume you've checked your connectivity? [03:13] Yes, I did and it's working fine [03:15] zeroseven0183: Can you branch other projects from launchpad? For example, lp:quickly? [03:15] On the same directory? [03:15] Ok, wait [03:16] Sadly, now [03:16] Same error "Permission denied (publickey). bzr: ERROR: Connection closed: Unexpected end of message. Please check connectivity and permissions, and report a bug if problems persist." [03:16] Hmmm...does bzr info output anything? [03:17] bzr: ERROR: Not a branch: "/home/tutdek/Quickly/". [03:17] zeroseven0183: OK - is there a .bzr subdirectory in that directory? [03:18] No, there is no .bzr [03:18] But what I see in my /home folder is .bazaar [03:20] zeroseven0183: Well if there's no .bzr directory then it's not even getting far enough to start creating the branch. The other possibility is permissions; maybe send off an email to the launchpad admins? [03:20] Alright but how come I was able to do that last night? [03:21] Something to do with my Public Key [03:22] I already generated and reposted my public key twice this morning [03:22] zeroseven0183: I don't know - at this point I'm speculating (note that the error message does say 'Permission denied' ). [03:22] Yup [03:23] zeroseven0183: Best bet is to contact the admins, they can probably tell you what's going on; maybe some spammer got your ISP subnet blacklisted. [03:24] The error means it's not succeeding at key auth. Which probably means the key you're using locally and the key on LP are different. [03:24] Try just connecting manually with ssh, and sprinkle on -v's to taste to see what key it's expecting to use. [03:25] I have two SSH keys in my Launchpad account [03:25] Are they both being scanned when branching? [03:26] I'd imagine. [03:26] But don't guess. See what ssh says. [03:28] Never thought the efects of cancelling that download would be harder [03:36] Alright, fullermd, how do I do the ssh -v thing? [03:36] Sorry I'm not good in command lines [03:36] zeroseven0183: "ssh -v bazaar.launchpad.net" [03:37] zeroseven0183: and see which keys it tries [03:37] The last line still says Permission denied (publickey) [03:38] May need a couple (-vv or -vvv) to see all the details; I don't know offhand what levels reports everything you need. [03:38] zeroseven0183: right, but hopefully the other lines will give you some information about which keys ssh is trying to use [03:43] Ok. Here's what I got http://pastebin.com/i17EkVBB [03:43] http://pastebin.com/raw.php?i=i17EkVBB [03:45] zeroseven0183: oh, I forget [03:46] zeroseven0183: you should also explicitly use your launchpad username with "ssh -v" [03:46] zeroseven0183: e.g. "ssh -vv bazaar.launchpad.net -l zeroseven0183" or whatever [03:46] (Unless your Launchpad username is tutdek?) [03:47] Nope [03:47] Otherwise that debug output just tells us that none of your local keys match the public key on launchpad for 'tutdek' :) [03:47] Here's the changes http://pastebin.com/raw.php?i=bEWWMRAd [03:48] Just added -l zeroseven0183 [03:49] Ok, then none of the keys you have locally in .ssh match any of the ones on your Launchpad account. [03:50] It says it is trying 3 different keys in your .ssh directory, but none of them are right. [03:52] zeroseven0183: take a look at your .ssh/id_rsa.pub file, does it match one of the ones at https://edge.launchpad.net/~zeroseven0183/+sshkeys ? [03:53] Now that's a lot of characters! :-) [03:54] Yes, I think they're match [03:54] Hmm! Then I'm not sure why it would fail :( [03:55] Wait [03:55] I have two keys in edge.launchpad.net [03:55] Yes. That's fine. [03:55] The last one matched with what's in the .pub [03:55] You can have twenty if you like ;) [03:56] Hmmm... I wish I was more patient last night [03:57] The download was just 50% when I cancelled it [03:57] I don't know why that key would have stopped working overnight. [03:57] This is the second time that a loggerhead bug has taken me weeks or months to track down and ended up being a tiny, tiny fix. [03:57] mkanat: heh [03:58] spiv: I guess I should go and e-mail the Launchpad admins? [03:58] zeroseven0183: at this point I only have increasingly unlikely theories, e.g. that the permissions on your .ssh directory or the .ssh/id_rsa file are wrong, but I'd expect ssh to warn if that's the case. [03:59] ("ls -ld ~/.ssh" should show "drwx------", and "ls -l ~/.ssh/id_rsa" should show "-rw-------") [04:00] Hmm: "debug2: we did not send a packet, disable method" is odd [04:01] :-( [04:02] It should have lines like "Offering public key: /home/andrew/.ssh/id_rsa", but it doesn't. [04:03] OK wait... I think I'm seeing some light [04:03] It definitely shows signs of being a local configuration problem, but I don't know what. [04:03] You mentioned the ls -l ~/.ssh/id_rsa [04:03] being in the ~/.ssh directory [04:03] I don't have that [04:03] Oh! I see. [04:03] So I moved the id_rsa and id_rsa.pub on that directory [04:04] I guess that's what the "((nil))" parts mean. [04:04] Actually, the filename was originally idko and idko.pub [04:04] then ran bzr branch lp:ubuntu-manual [04:04] * fullermd blinks. [04:04] You can set the filename in your .ssh/config. [04:04] Not that there's much reason to use anything other than the default if you only have one key on that system. [04:04] And now it's trying to fetch the revisions [04:05] Yup [04:05] Ah good. Glad I could help! [04:05] I think I can breathe now [04:05] Sure [04:05] Everyone who chatted with me here were very helpful [04:05] What a learning experience [04:06] Thank you very much, people [04:06] Well, you've reminded me of some details of how to interpret ssh -vv output :) [04:07] And now for the most exciting part of the day... LUNCH! [04:07] I'm off now, guys [04:07] Thank you very very much! [04:07] You did very well in helping me [04:07] See you around next time [04:07] :-) [04:17] mwhudson: Okay, MP: https://code.launchpad.net/~mkanat/loggerhead/limit-pygments/+merge/23900 [04:20] mkanat: if (): looks like C :) [04:20] mwhudson: Oh, right. I'm sleepy. :-) [04:49] * spiv -> lunch [06:16] * man_in_work waves [06:18] i'm trying to branch a bzr repo and i'm getting "No repository present" [06:19] using command: bzr branch sftp://user@server/path/to/branch/ [06:20] /path/to/branch exists as an absolute path on the server, and includes a .bzr directory [06:20] inside that, there is a file "branch-format", which reads: Bazaar-NG meta directory, format 1 [06:21] my bzr --version reads: Bazaar (bzr) 2.1.1 [06:21] The whole output would be helpful. Also an 'info' of that URL. [06:22] bzr: ERROR: No repository present: "sftp://user@server/path/to/branch/" [06:22] ^^ whole output [06:22] what do you mean by info? [06:22] bzr info $URL [06:22] same output [06:23] What all is in the .bzr? [06:24] branch-lock, branch directories, branch-format and README files [06:26] is bzr-ng a different tool to bzr 2.x? [06:26] bazaar-ng is bzr. I'm not sure what bzr-ng is. [06:26] What about in /path/to and /path? That doesn't list a repository. [06:29] there is a .bzr in the 2nd-level directory in /path/to [06:29] Does the user have access to it? [06:31] perms are 755 on directories and 644 on files [06:31] so, yes [06:31] Well, bzr is telling you that for some reason it can't get its hands on it. [06:31] Maybe .bzr.log says something? There's some -D flag to get it to tell more about what it's doing... [06:33] -Dsftp maybe? Though I suspect that talks more about the interaction with sshd itself than the activity happening over it. [06:33] I thought there was a -Dtransport or something, but I don't see it in the docs. [06:33] File "/usr/lib/python2.5/site-packages/bzrlib/bzrdir.py", line 684, in find_repository [06:33] raise errors.NoRepositoryPresent(self) [06:33] ^^ that's the end of the backtrace in the log [06:33] If you've got bzr on the server, bzr+ssh has some more options (and is a better all around choice anyway) [06:33] Yeah, but that doesn't tell us where it {was looking,had looked}. [06:34] some smart fellow decided to use a bzr repo on a server that doesn't have bzr installed [06:34] don't ask me why [06:35] Well, you could try pretending to be bzr. Use sftp(1) yourself to go in as that user and poke around, make sure you can reach into the repo. [06:36] spiv: ^^ am I missing a -D option to tell what files it's trying to look up? [06:36] don't see any problems [06:36] with direct sftp access [06:38] i don't see -D on the manpage anywhere [06:38] is it a flag to bzr or to branch? [06:38] To bzr. [06:38] `bzr help debug-flags` [06:46] igc, source is properly frozen now [06:46] https://edge.launchpad.net/bzr/2.2/2.2b2 [06:47] poolie: Oh, you're around to bug instead of spiv :) [06:48] poolie: thanks. I'll check the installer builds now [06:48] poolie: Am I misremembering that there used to be a -Dtransport or the like? [06:48] not for much longer :) [06:48] fullermd: there is log+sftp:// etc [06:48] i would like a debug option that puts this onto all transports but i don't think it exists yet [06:48] Ooh, maybe that's what I'm trying to think of. [06:48] man_in_work: Try that ^^^ [06:49] got some useful info [06:51] No such file: '/path/to/branch/.bzr/checkout/format': [Errno 2] No such file [06:51] Probably irrelevant. 'branch' doesn't really do anything meaningful with that. [06:53] what about repository/format? [06:55] That's more what we're be looking for. [06:56] well, that also doesn't exist [06:56] there's no repository directory [06:57] i see a repository.backup in an ancestor directory's .bzr [06:58] looks like someone's been screwing around [07:07] fullermd: thanks for your help [07:36] ping: vila [07:36] naoki: yeah ! pong [07:36] vila: https://code.launchpad.net/~songofacandy/bzr/fix-523746-dev/+merge/20537 [07:36] naoki: yup, can you break it ? [07:37] vila: bzr uses win32utils.get_unicode_argv. So bzr recieve args by unicode [07:40] naoki: meh [07:40] _o/ [07:40] _o| [07:40] _o/ [07:41] Please wait some minutes. (I'l [07:42] I'll test your branch from now. [07:43] I'm using Japanese WinXP. [07:43] add new file in Chinese name. commit, and modify it. [07:45] vila: http://paste.ubuntu.com/420255/ [07:45] This is your fix. [07:46] naoki: great, can you send me a tarball of the wt/branch/repo [07:47] what is cp932 ? [07:47] Japanese codepage. It's Shift_JIS + some extra chars. [07:48] is the filename valid in this encoding ? [07:48] No. It isn't. [07:48] then we have a far larger problem [07:48] But I can add that file name from TortoiseBZR, qbzr, etc. [07:49] Because bzr recieves args by unicode. [07:49] I'm trying to convert a repository stored on a remote server to which I have only SSH and SFTP access; it doesn't have Bazaar installed. [07:49] $ bzr check sftp://bzr@fs/~/ [07:49] Format RepositoryFormatKnit3() for sftp://bzr@fs/~/.bzr/ is deprecated - please use 'bzr upgrade' to get better performance [07:49] naoki: from the command line ? [07:49] $ bzr upgrade sftp://bzr@fs/~/ [07:49] bzr: ERROR: Cannot convert from format RepositoryFormatKnit3() to format RepositoryFormat2a(). Does not support nested trees [07:50] am I correct in thinking that current Bazaar 2.1 should be able to convert any previous Bazaar repository format to 2a? [07:50] if not, what are the limits? if so, why is this not working? [07:50] vila: I'ts very difficult. I can't input chinese from keyboard. But I can copy&paste the filename. [07:51] vila: I can do diff in cmd.exe http://paste.ubuntu.com/420255/ [07:52] bignose: Knit3 is dirstate-with-subtree, you can only convert it to other formats which also support -subtree. That's what the "Does not support nested trees" is about [07:52] naoki: 'bzr diff' alone works right ? [07:53] vila: filename is broken. But others are all right. [07:53] naoki: I smell something wrong about bzr receiving unicode... [07:53] naoki: broken how ? [07:54] naoki: is GetCommandLineW documented somewhere ? My feeling is that this function handle the conversion from the input encoding to unicode [07:55] vila: This is how to reproduce http://paste.ubuntu.com/420255/ [07:55] naoki: well, almost, I don't have winmerge nor your branch :-/ [07:56] vila: No. On Windows, filenames and commandline arguments are unicode internally. [07:56] vila: So, sys.argv is encoded string. Not original. [07:56] naoki: yes, in bzr too, yet there are encoded.edecoded at appropriate times [07:56] you mean decoded ? [07:57] python use decode to go from one encoding to unicode and encode to go from unicode to one encoding [07:57] maxb: okay. so what current format could I convert this repository to? [07:57] vila: I means Python's default sys.argv. It is from main(int argc, char *argv[]). And argv is encoded. [07:58] (Python3's sys.argv is from main(int argc, wchar *argv[]). So it's native unicode) [07:59] ‘bzr help current-formats’ doesn't mention subtree anywhere in its list. [07:59] is there any currently-supported format that supports subtree? [07:59] naoki: AFAIK you can't *type* unicode on the command line, you always use the user encoding *then* python decodes it into unicode [08:01] vila: Windows command prompt is based on unicode. So I can use unicode on command line. [08:01] * vila scratches head [08:01] naoki: it's important to realise that “encoded string” is *not* Unicode. Those are two different data types. [08:02] either you have an encoded string, which is not unicode [08:02] vila: And win32utils.get_unicode_argv() recieves command line by unicode. [08:02] or you have unicode, which is not an encoded string. [08:02] you need to be clear about which you have at any given moment. [08:03] naoki: win32utils.get_unicode_argv() relies on GetCommandLineW and I suspect the later handles the decoding, where is it documented ? [08:04] vila: commandline is UTF-16 in Windows. GetCommandLineW gets the commandline without encoding/decoding. [08:04] * vila blinks [08:05] naoki: there are several places in bzr where we use osutils.get_user_encoding() even on win32 (see win32utils), I'd really like to understand what is happening here [08:05] bialix: can you help here ? [08:06] bzr uses that encoding when output to console. [08:07] And then, console decode it to utf-16 and show. [08:08] On windows API, everything is Unicode based. And encoded string is used for compatibility to ascii based application. [08:08] naoki: no, we use different encodings for: use, terminal and file system [08:09] http://stackoverflow.com/questions/846850/how-to-read-unicode-characters-from-command-line-arguments-in-python-on-windows seems to imply that GetCommandLineW is indeed decoding into unicode [08:10] naoki: i.e. the command-line itself is not in unicode, it's decoded by GetCommandLineW [08:11] vila: GetCommandLineW doesn't decode anything. commmandline is unicode and GetCommandLineW pass it. [08:13] naoki: why does python needs GetCommandLineW then ? [08:13] vila: sys.argv is **encoded** arguments. So it may be broken. [08:14] vila: It is why we should use GetCommandLineW to get unicode arguments. [08:15] naoki: ok, so we disagree on that, you say the command line is unicode, I say it's encoded in user encoding, what code can you use to decide on that ? [08:15] naoki: 'sys.argv is **encoded**' by who ? [08:16] naoki: and how ? That's what we need to find to be able to replicate around subprocess.Popen [08:16] vila: MSVCRT encode arguments and pass it to main() [08:16] naoki: then Popen most probably will do the same [08:16] vila: Python3's subprocess module uses CreateProcessW. So it can pass unicode arguments to child process. [08:17] vila: But Python2's subprocess can't handle unicode. [08:17] haaaaa, but we are stuck with python2 so far [08:17] naoki: can we trick it into accepting utf8 encoding ? [08:17] vila: So, there are 2 ways: 1. Use CreateProcessW manually like GetCommandLineW. [08:18] vila: on Windows command line is unicode [08:18] vila: 2. replace chars that doesn't represent in user_encoding. [08:19] bialix: meh, when is the user encoding used then ? [08:19] naoki: 1) sounds better, [08:20] vila: internal Windows API is unicode (functions end with W), but there is also ANSI API (functions end with A), under the hood A functions do encoding on the fly, IIUC [08:20] naoki: about 2) replacing chars means losing some info, what are we trynig to preserve here and can't we switch to a purely arbitrary name with ascii only chars ? [08:20] python uses A [08:21] I guess for Windows 98 compatibility [08:22] Hi guys! I've just offered the bzr-bash-completion plugin for merging into bzr.dev, as distro packagers as well as users mostly prefer this over an independent plugin. https://code.launchpad.net/~gagern/bzr/bug560030-include-bash-completion-plugin/+merge/23912 open for comments. I'll be around. [08:24] bialix, naoki: Sorry to sound dense, but how to you guys can be sure that the command line is unicode ? The fact windows *internally* uses a unicode API doesn't guarantee that the command line itself doesn't respect the user encoding [08:24] otherwise what's the point of the user encoding ? Why do we use it in win32utils.get_appdata_location() for example ? [08:24] vila: it depends on how program is launched [08:25] bialix: all right, let's focus to subprocess.Popen then since that's where the bug seems to be actually [08:25] user encoding? what's wrong with user encoding? [08:25] vila: subprocess.Popen can't use unicode API [08:26] there is some code in subprocess.py explicitly disabled unicode on Windows. [08:26] bialix: nothing wrong with user encoding, my question is why do we use it to *decode* some data if it's already unicode ? [08:26] vila: lemme look at get_appdata...blah-blah-blah [08:27] vila: I don't see there user_encoding [08:27] naoki: to be clear, I'm not saying you're wrong, just that I don't understand the logic here :) [08:28] bialix: wtf [08:28] vila: os.environ holds data as plain strings [08:28] bialix: oh, it's in the docstring [08:29] Mark Hammond said it's actually should be 'mbcs' encoding [08:29] bialix: hmm, mbcs is the file system encoding right ? [08:30] well, mbcs it's a ... mbcs! [08:30] file system encoding depends on file system. AFAIK NTFS s unicode [08:31] bialix: ha, look at _ensure_unicode which call get_user_encoding [08:31] I'm not quite understand internals of mbcs, but it's a very strange thing [08:31] vila: yep [08:32] vila: on my Russian system mbcs == cp1251 [08:32] cp1251 is also russian [08:32] maybe on Japanese/Chinese systems mbcs means something different, I really don't know [08:34] 'mbcs' is an alias. It is set by site.py [08:35] I suspect something like that [08:41] spawn is easier than CreateProcess: http://paste.ubuntu.com/420255/ [08:41] naoki: I just pushed a new revision (forcing mbcs for subprocess.Popen), can you re-try [08:42] naoki: that's still the same url you've been pasting for the third time, is this intended ? [08:42] miss. http://paste.ubuntu.com/420272/ [08:43] naoki: that one is about spawn [08:44] spawnw uses unicode. [08:44] So I think we can use it instead of subprocess.call [08:44] AFK, sorry. [08:45] naoki: you mean _wspanlp uses unicode, ok, but what was 'miss' referring to ? === ctrlsoft is now known as jelmer [09:42] its kindof sad shutting down my persistent IRC server [09:42] Heh. [09:43] Try to think of it as liberating! [09:43] sure [09:43] * lifeless tries [09:43]