| spiv | mwhudson: regarding python backtraces from gdb, have you seen https://fedoraproject.org/wiki/Features/EasierPythonDebugging ? | 00:09 |
|---|---|---|
| RenatoSilva | ok never mind I'm writing a script and I'm cd'ing to the dir anyway | 00:09 |
| RenatoSilva | thanks all | 00:09 |
| mwhudson | spiv: no i haven't | 00:10 |
| mwhudson | spiv: rather a long way away from being supported in the gdb in hardy, i'd expect... | 00:11 |
| spiv | mwhudson: probably | 00:12 |
| poolie | spiv, did you see https://code.launchpad.net/~gz/bzr/test_ssh_client_medium_eintr__read_bytes/+merge/19439 | 00:14 |
| spiv | poolie: not yet, will look tomorrow | 00:22 |
| poolie | cool | 00:22 |
| poolie | how is V? | 00:22 |
| spiv | Sleeping at the moment, thankfully. He was up late last night! | 00:23 |
| poolie | he looks just like you :-P | 00:30 |
| Noldorin | can i force a push to a bzr branch on launchpad that has diverged from the local one? | 00:47 |
| lifeless | yes | 00:48 |
| bob2 | bzr help push -> overwrite, beware the consequences | 00:48 |
| === jamesh_ is now known as jamesh | ||
| Noldorin | bob2, yup. i just managed to clobber the launchpad branch, so no worries. thanks | 00:49 |
| AfC | Noldorin: I think bob2 more means "consequences" more means "if anyone has grabbed the [revisions on the] previous version of the branch, then you'll never be able to get rid of them (should that person subsequently submit a patch to you and/or merge from your new history)" | 00:54 |
| Noldorin | AfC, oh, fair enough. fortunately i'm the only dev on the project at the moment, and no-one else is submitting patches | 00:55 |
| Noldorin | but i see the point | 00:55 |
| fullermd | In those situations I like to use profanity. | 00:55 |
| Noldorin | fullermd, hehe | 00:55 |
| * fullermd watches openoffice build all his disk space away. | 00:57 | |
| AfC | I gotta say, I don't miss rebuilding all of oo for a single line patch | 00:59 |
| poolie | lifeless: does https://bugs.launchpad.net/bugs/522637 ring any bells? | 00:59 |
| ubottu | Ubuntu bug 522637 in bzr "BzrCheckError: Cannot add revision(s) to repository: missing referenced chk root keys" [High,Confirmed] | 00:59 |
| fullermd | Pfft, the warning says it eats 11 gig of space for the build. But it's barely 10.5; getting me all worried about nothing... | 01:00 |
| fullermd | And to think, earlier I actually had a momentary thought of "Gee, 6.5 megs... that seems a big tarball for a VCS" | 01:01 |
| spiv | poolie: oh hey, that's some of that paranoia I added just before 2.0, I think? Nice (relatively speaking) to see it catch something :) | 01:01 |
| lifeless | poolie: yes, isn't there anothe rbug open like that? be great to help the user, figure it out and fix it. | 01:02 |
| * igc lunch | 01:50 | |
| parthm | Hello, I am trying to understand bzr test infrastructure better in the context of http://old.nabble.com/Re%3A-help-needed-in-fixing-test-failure-%28-138600%29-p27617476.html | 04:09 |
| parthm | after doing self.make_branch('foobar') I drop into pdb | 04:09 |
| parthm | doing a system('ls | 04:09 |
| parthm | doesnt show any directories. I am in test_home_dir. Should I be seeing a foobar dir? how does that work? | 04:10 |
| mwhudson | parthm: which testcase class did you inherit from? | 04:13 |
| parthm | TestCaseInTempDir | 04:14 |
| parthm | I am trying to add the test case to blackbox/test_versioning.py: TestVersioning | 04:14 |
| mwhudson | parthm: make_branch in that class only writes to a memory-backed transport, not the filesystem | 04:15 |
| parthm | Ah. OK. Basically what I need to do is create a directory thats not a part of any repo/branch. I need to run 'bzr mkdir abc' there and ensure that dir abc is not created. | 04:17 |
| parthm | What would be a good way to do that considering that test_home_dir is already a branch | 04:17 |
| mwhudson | not sure tbh | 04:19 |
| mwhudson | poolie: can you help parthm > | 04:19 |
| mwhudson | ? | 04:19 |
| poolie | hi there | 04:19 |
| poolie | yes probably | 04:19 |
| poolie | on the phone, will just be a minute | 04:20 |
| mwhudson | parthm: if you inherit from TestCaseWithTransport, then self.make_branch _will_ create a directory, which might make things easier to understand | 04:20 |
| parthm | hi poolie | 04:20 |
| mwhudson | might not be the right thing to do though | 04:20 |
| parthm | That seems fine (with my limited understanding). I am looking at blackbox.test_status.BranchStatus.test_tree_status_ignores. Maybe I can do a make_repository('foo', shared=True). | 04:27 |
| poolie | (back) | 04:27 |
| poolie | so istm that you have to use a real directory | 04:27 |
| parthm | If this creates a dir. the mkdir should fail in this case also. | 04:27 |
| poolie | at the moment wt must go to disk | 04:27 |
| parthm | Yes. | 04:28 |
| parthm | The 'init-repo' failcase http://pastebin.com/m6e5397c9 | 04:29 |
| * poolie looks | 04:31 | |
| poolie | ok | 04:31 |
| poolie | and you're having trouble reproducing that inside a test? | 04:31 |
| poolie | rebooting, biab | 04:32 |
| parthm | Yes. The same failure can also be seen on the command line with a simple directory thats not a branch. I am trying to create the test case for https://bugs.launchpad.net/bzr/+bug/138600 | 04:33 |
| poolie | yep | 04:33 |
| poolie | i saw your mail earlier | 04:33 |
| poolie | so what happens? | 04:33 |
| ubottu | Ubuntu bug 138600 in bzr "`bzr mkdir` create new dir before it checks is branch/checkout exists" [Low,In progress] | 04:33 |
| poolie | it still sees the directory even with your fix in place? | 04:33 |
| parthm | Really? Are you trying this branch https://code.launchpad.net/~parthm/bzr/138600 are are you referring to the pastebin | 04:34 |
| parthm | The pastebin is bzr 2.0.4 | 04:35 |
| poolie | no, i'm asking where you're getting stuck now | 04:35 |
| parthm | When I run bzr from the command line the fix seems to work. I am trying to add the test case right now. | 04:36 |
| poolie | ok | 04:36 |
| poolie | so i think the approach you described should work | 04:36 |
| poolie | make a bzr repository, then try bzr mkdir inside that | 04:37 |
| parthm | OK. The TestCaseWithTransport approach worked. Thanks mwhudson, poolie. | 04:47 |
| parthm | One more question. Unlike other commands, mkdir doesn't seem to be having blackbox.test_mkdir. All tests are in blackbox.test_versioning.TestVersioning, so should I just add blackbox.test_versioning.TestMkdir or should it be blackbox.test_mkdir.TestMkdir? | 04:49 |
| parthm | Hi poolie. I was just mentioning that the TestCaseWithTransport approach worked. | 04:49 |
| poolie | oh ok | 04:49 |
| poolie | so you're all good now? | 04:49 |
| parthm | Yes. Just one more question. Unlike other commands, mkdir doesn't seem to be having blackbox.test_mkdir. All tests are in blackbox.test_versioning.TestVersioning, so should I just add blackbox.test_versioning.TestMkdir or should it be blackbox.test_mkdir.TestMkdir? | 04:50 |
| poolie | probably add a new module for consistency | 04:50 |
| poolie | but either is fine | 04:50 |
| parthm | OK. | 04:50 |
| parthm | I will add the tests and send a merge request for bug 138600. | 04:51 |
| ubottu | Launchpad bug 138600 in bzr "`bzr mkdir` create new dir before it checks is branch/checkout exists" [Low,In progress] https://launchpad.net/bugs/138600 | 04:51 |
| parthm | Does the NEWS update need to be a part of the merge request or is that a part of the landing process? | 04:51 |
| poolie | it's a bit better to put it in your proposal so that we can review that too | 04:54 |
| poolie | also it acts as a good summary of what you think you've changed | 04:55 |
| parthm | OK. Will do. Thanks. | 04:55 |
| poolie | wow the review queue is so long | 04:56 |
| spiv | Yeah wow, lots of reviews. I guess I know what I'm doing tomorrow then :) | 05:02 |
| * spiv wanders off again | 05:03 | |
| parthm | Is there a good way to check (and exit) if the current dir is a workingtree? currently I am doing "wt, dd = WorkingTree.open_containing(d)" | 05:06 |
| poolie | that's probably good | 05:08 |
| poolie | if you want to know about the specific directory, not just a containing directory, just use WT.open() | 05:08 |
| poolie | bzrdir may have a method to tell you? | 05:08 |
| parthm | The problem I am seeing with the current fix is that if we have abc/xyz where abc is versioned and xyz is unknown, if user does 'bzr mkdir ./xyz/foo', 'foo' is created before bzr exists. So I suppose I need to know if the first part of the path is versioned. | 05:10 |
| poolie | oh i see | 05:12 |
| poolie | right, that seems like a separate bug | 05:12 |
| poolie | can you see how to reproduce this? | 05:13 |
| parthm | Yes. I added a test case where I do os.mkdir('./abc') and chdir('./abc'). run_bzr('mkdir xyz') would create xyz. | 05:14 |
| poolie | k | 05:15 |
| poolie | biab | 05:15 |
| igc | bbiab | 05:25 |
| poolie | cheerio | 05:25 |
| parthm | for some reason "print" in builtins.cmd_xxx cause a UnicodeEncodeError while running selftest. If I run bzr from the command like this works ok. I tried this with cmd_status and cmd_mkdir while debugging mkdir. Is this expected? | 05:49 |
| bob2 | printing unicode strings works differently when stdout is not a terminal | 05:49 |
| bob2 | maybe that's what you're encountering | 05:49 |
| poolie | parthm: yes, the difference is that it's probably going into a stringio | 05:50 |
| poolie | are you printing something nonascii? | 05:50 |
| parthm | No. Even "print 'hello'" fails. | 05:51 |
| poolie | !! | 05:51 |
| parthm | http://pastebin.com/m51e779bc | 05:52 |
| parthm | Whats the suggested way to put a debug print? | 05:52 |
| parthm | Never mind. I will explore some more. Seems to be some issue in my setup. Even modifying an existing outf.write seems to be failing | 05:58 |
| poolie | hi | 05:59 |
| poolie | that's strange | 05:59 |
| poolie | parthm: if you want to get arountd it | 05:59 |
| poolie | at least for now | 05:59 |
| poolie | change the note() code to do | 05:59 |
| poolie | msg.encode(self.stdout.encoding, errors=replace) | 06:00 |
| poolie | but this is a bit strange | 06:00 |
| parthm | So I need to put the msg.encode(..) in my run? | 06:00 |
| poolie | in trace.note | 06:14 |
| poolie | i'm just testing your permissions patch | 06:14 |
| poolie | thanks for doing that | 06:14 |
| poolie | much karma for fixing elmo's bugs :) | 06:15 |
| poolie | parthm: stuck? | 06:21 |
| parthm | Thanks :). I am trying to figure out what might be a good fix for http://pastebin.com/m56f77879 case. Switching mkdir with open_containing (http://pastebin.com/m2648b1fb) works for the repo case or when user just does a 'mkdir xxx' but this case still fails. | 06:27 |
| poolie | parthm: i think run_bzr has a parameter to set the cwd so you don't need to manage it yourself | 06:29 |
| poolie | or use a scriptrunner | 06:30 |
| parthm | Yes. Will do that. | 06:30 |
| poolie | this is the thing about making sure it fails if the parent is not versioned? | 06:31 |
| poolie | can you add it to the inventory before it exists? | 06:31 |
| poolie | it seems like that would be cleanest | 06:31 |
| parthm | I don't think I understand. So if user does a mkdir xxx/yyy and xxx is not versioned (unknown) whats a good way to check and fail? | 06:33 |
| poolie | parthm: what i'm suggesting is that we try to add yyy to the inventory | 06:34 |
| poolie | ie to the metadata | 06:34 |
| poolie | before making the directory | 06:34 |
| poolie | that will fail if xxx is not already in there | 06:34 |
| poolie | then if that succeeds, mkdir | 06:34 |
| poolie | the only catch will be if the add-to-inventory code checks it exists on disk | 06:35 |
| poolie | which would be a bug imo, but maybe a slightly harder one | 06:35 |
| parthm | Ah. I get it. What API should I look at? | 06:35 |
| poolie | basically i'm just suggesting moving the mkdir after wt.add | 06:37 |
| poolie | it may fail horribly | 06:37 |
| poolie | but it's worth a go | 06:37 |
| parthm | So this is what my current cmd_mkdir looks like http://pastebin.com/m57cffad5 . unfortunately I hit this painful unicode error http://pastebin.com/m3b815639 . I am using vim which I think handles unicode fine. Wow. | 06:45 |
| poolie | ouch | 06:47 |
| poolie | did you try changing note() to print the message with errors='replace' rather than spazzing out? | 06:47 |
| poolie | i think you need to get unblocked from that to make other changes easer | 06:49 |
| poolie | i don't know why it's failing there | 06:49 |
| parthm | Yes. I just did the note change. That worked, thanks. | 06:49 |
| poolie | it is annoying certainly | 06:49 |
| parthm | http://pastebin.com/m57cffad5 doesn't seem to work. I may be doing something wrong. | 06:50 |
| poolie | what happens? | 07:17 |
| poolie | it's unfortunate you have to add the root | 07:17 |
| poolie | though that is a bit of a separate issue | 07:18 |
| parthm | Looks like the msg.encode change didn't work. msg is undefined. http://pastebin.com/m579486a7 Sorry I seem to have more questions than answers :( So is note called by self.outf.write? | 07:21 |
| parthm | Can wt.has_filename be used for a conditional check on os.path.abspath(base)? We need to use abspath as the user may either do 'bzr mkdir foo/abc' or 'cd foo; bzr mkdir abc | 07:24 |
| parthm | where foo is unversioned | 07:24 |
| poolie | no, note should not be called by self.outf.write | 07:24 |
| poolie | parthm: when you open the wt with open_containing, it gives you back the relative path | 07:25 |
| poolie | you can then chop off all but the last component of that and check that it's versioned | 07:25 |
| parthm | Oh OK. That makes sense. | 07:25 |
| parthm | What do you suggest I do about the prints for now? | 07:26 |
| poolie | which prints are these? | 07:32 |
| poolie | just ones you're adding for debugging? | 07:32 |
| parthm | Yes. | 07:34 |
| parthm | This is the final version of cmd_mkdir.run http://pastebin.com/m3893f263 All tests are passing, yay! | 07:35 |
| parthm | These are the three test cases | 07:36 |
| parthm | http://pastebin.com/m2e37b536 | 07:36 |
| === radoe_ is now known as radoe | ||
| poolie | i think you need better docstrings | 07:53 |
| poolie | rather than 'invalid' say 'unversioned' | 07:54 |
| parthm | I agree. Will do that. | 07:54 |
| parthm | OK. | 07:54 |
| poolie | and rather than joining the strings, use ['mkdir', newdir] | 07:54 |
| poolie | thanks for all the patches | 07:54 |
| parthm | Welcome :) | 07:54 |
| parthm | I am trying the winepython usage you suggested. | 07:55 |
| parthm | After installing testtools and paramiko I get "bzr: ERROR: No module named Crypto.Util.randpool" | 07:55 |
| parthm | Any ideas? | 07:55 |
| parthm | paramiko installed without any errors. | 07:56 |
| poolie | blah | 08:10 |
| poolie | it's some kind of known bug about paramiko | 08:10 |
| poolie | just run particular tests and avoid paramiko | 08:11 |
| poolie | it's not very relevant to your bugs | 08:11 |
| poolie | well, it's time for me to go | 08:13 |
| poolie | push your branch and we'll rereview it | 08:13 |
| parthm | OK. Bye. | 08:14 |
| lvh | where can I see the branch format of a branch on launchpad? | 08:20 |
| lvh | I just branched something using 2.0.4 and it was hella slow, AFAIK the main cause for that is typically branch format mismatch | 08:20 |
| lvh | Ack. Why do I always spend 20 minutes looking for something, ask for help, and then find it immediately after? :| | 08:20 |
| lvh | Sorry, I'm having to deal with a really bad wifi connection | 08:37 |
| lvh | I've got a format 7 branch off of a format 7 trunk | 08:38 |
| lvh | the format 7 trunk is synced from svn, it's not actually pushed to | 08:38 |
| lvh | what are my options for upgrading the local format ? | 08:38 |
| jderose | lvh: `bzr upgrade` should upgrade to local repo to the 2.0 format | 09:23 |
| GaryvdM | Hi igc | 10:19 |
| GaryvdM | igc: I'm trying to find the page that you had the OOo branch on. | 10:20 |
| igc | hi garyvdm | 10:20 |
| igc | garyvdm: if you want some large test repos, try http://people.canonical.com/~ianc/benchmarks/repos/bzr/ | 10:21 |
| GaryvdM | Thanks | 10:23 |
| GaryvdM | Hmm 1.3 gB. That will cost me about R 260 to download :-( | 10:23 |
| GaryvdM | ok - bye | 10:25 |
| AntoineLaf | I have a very beginner level question and hopefully someone would be kind enough to point to where I could find some answers: here it goes. I'm trying to make sense of the "markers (not sure if it's the properword)" inserted in a conflicting merged file. I would like to know where I could find an explanation for <<<< TREE and >>>>>> MERGE SOURCE. Thank you. | 10:40 |
| Kinnison | IIRC what you get is: | 10:40 |
| Kinnison | <<<< TREE | 10:40 |
| Kinnison | what you had in your tree | 10:41 |
| Kinnison | so this is your change | 10:41 |
| Kinnison | ======= | 10:41 |
| Kinnison | what the merge source had which conflicted | 10:41 |
| Kinnison | this is what you're trying to merge in from the remote branch | 10:41 |
| Kinnison | >>>>>> MERGE SOURCE | 10:41 |
| Kinnison | type thing | 10:41 |
| Kinnison | does that help? | 10:41 |
| AntoineLaf | so what is before the ====== part is the changed part | 10:42 |
| AntoineLaf | and under the ====== is the part I try to merge and is conflicting... | 10:43 |
| AntoineLaf | I think it helps, let me have a better look at it again | 10:43 |
| * Kinnison nods | 10:43 | |
| AntoineLaf | <<<<<<< TREE | 10:45 |
| AntoineLaf | something | 10:45 |
| AntoineLaf | ======= | 10:45 |
| AntoineLaf | >>>>>>> MERGE-SOURCE | 10:45 |
| AntoineLaf | something like that means that the merge tries to remove something? | 10:45 |
| === salgado-afk is now known as salgado | ||
| Kinnison | It means that where the local branch changed <foo> to "something", the merge branch removed it | 10:48 |
| Kinnison | yes | 10:48 |
| AntoineLaf | this is starting to make sense... | 10:48 |
| AntoineLaf | I guess I'll get more scout badges this month... | 10:49 |
| AntoineLaf | :) thx | 10:49 |
| Kinnison | :-) | 10:49 |
| Kinnison | Scouts have a DVCS badge? | 10:49 |
| AntoineLaf | yes, with 20 levels with different colors... | 10:50 |
| AntoineLaf | sorry, this is my fantasy world... | 10:50 |
| AntoineLaf | have to say that bazaar is the first versioning system that I finally had the guts to stick with and learn... I'm from a designer's background... not all that familiar with the thinking behind the tool. anyways. | 10:51 |
| === AnMaster_ is now known as AnMaster | ||
| bialix | what is the new feature in bzr? cmd_foo.run_direct? | 11:14 |
| igc | night all | 11:39 |
| radoe | Hm, will there be full releasenotes for 2.1 put online? By now it seems one has to check all 2.1.0bX and 2.1.0.rY together with 2.1.0 to get all incremental changes. | 12:17 |
| adiroiban | hi. I'm using bzr break-lock , but it tell me that the lock is in used and I can not break it | 12:25 |
| adiroiban | is there a bzr command to show who is using t he lock? | 12:25 |
| bialix | adiroiban: break-lock actually show you info about lock holder, do you have different experience? | 12:35 |
| bialix | this command is interactive: it show you info about lock and then asking for y/n answer | 12:36 |
| adiroiban | http://paste.ubuntu.com/378311/ | 12:36 |
| adiroiban | this is all I see | 12:36 |
| adiroiban | no prompt for breaking the lock | 12:36 |
| adiroiban | or hint about how can I solve this issue :) | 12:37 |
| bialix | adiroiban: please run again as `bzr -Derror break-lock` and pastebin | 12:40 |
| bialix | I suspect some application holds OS lock on dirstate file | 12:41 |
| adiroiban | http://paste.ubuntu.com/378312/ | 12:41 |
| adiroiban | ..maybe, I'm not greping lsof | 12:42 |
| bialix | most likely some application still has OS lock over file .bzr/checkout/dirstate | 12:43 |
| bialix | usually you will see different message | 12:43 |
| bialix | or there is bug | 12:44 |
| adiroiban | yep. lsof, shows that gedit is using it (I'm using the gedit bzr integration) | 12:47 |
| adiroiban | I still think it is a bug as bzr should inform me that gedit is messing the repository | 12:47 |
| bialix | adiroiban: I don't know is there any API for this | 12:50 |
| bialix | but I think you need to file a bug report | 12:50 |
| === mrevell is now known as mrevell-lunch | ||
| === mrevell-lunch is now known as mrevell | ||
| === beuno_ is now known as beuno | ||
| === salgado is now known as salgado-lunch | ||
| === salgado-lunch is now known as salgado | ||
| === salgado is now known as salgado-brb | ||
| Tak | hmm, should bzr-svn be able to auto-authenticate me over svn+https? | 15:18 |
| jelmer | Tak: what do you mean by auto-authenticate? | 15:20 |
| Tak | I mean, use my credentials from authentication.conf | 15:21 |
| jelmer | Tak: it should be able to use a username/password if one is set in ~/.subversion or ~/.bazaar/authentication.conf | 15:21 |
| jelmer | but you need a recent version for ~/.bazaar/authentication.conf to work | 15:21 |
| Tak | ok | 15:21 |
| Tak | I have bzr 2.0.4 and bzr-svn 1.0.2 | 15:21 |
| jelmer | that should be recent enough | 15:22 |
| Tak | oh, what should the scheme be? | 15:22 |
| jelmer | oh actually, authentication.conf perhaps doesn't work for svn+https, only https | 15:22 |
| Tak | I have svn+https | 15:22 |
| jelmer | for any particular reason? just https should work.. | 15:24 |
| Tak | it wasn't working for me at some point, with this server | 15:24 |
| Tak | but that was several versions ago; let me try again | 15:24 |
| Tak | ok - it looks like it tried to authenticate me that time - are there character restrictions on the passwords in authentication.conf? | 15:28 |
| jelmer | Tak: it should be a utf8 string | 15:32 |
| Tak | ok, but it can contain *&^#$^$# ? | 15:32 |
| Peng | The config parser _might_ get a bit confused. E.g., # denotes comments. Try putting it in quotes? | 15:34 |
| Peng | It's mostly standard ConfigObj, FYI. | 15:34 |
| Peng | If you like using the Python prompt, import bzrlib.util.configobj.configobj and check how it parses it. | 15:35 |
| Tak | hmm, quotes don't help | 15:35 |
| * Tak do that | 15:35 | |
| Tak | yeah, it's truncating the password at the comment char | 15:40 |
| Tak | if I single-quote it, it looks ok in the ConfigObj, but the authentication still fails | 15:42 |
| Tak | is it maybe getting parsed again someplace else? | 15:43 |
| === salgado-brb is now known as salgado | ||
| bialix | http://martinfowler.com/bliki/VersionControlTools.html | 16:00 |
| bialix | bzr is not mentioned at all | 16:00 |
| bialix | maybe somebody know why? | 16:00 |
| ronny | bialix: he write why in the text | 16:03 |
| ronny | -i+o | 16:03 |
| bialix | hmm, I've missed it | 16:04 |
| bialix | ok, found | 16:04 |
| === beuno is now known as beuno-lunch | ||
| === beuno-lunch is now known as beuno | ||
| bjp | i'm seeing this bug: https://bugs.launchpad.net/loggerhead/+bug/518679 | 18:06 |
| ubottu | Ubuntu bug 518679 in loggerhead "Unable to see diffs" [Undecided,Incomplete] | 18:06 |
| bjp | with bzr version 2.1.0dev5 | 18:06 |
| bjp | and loggerhead from bzr reversion 400 | 18:06 |
| bjp | though the second to last comment implies this is already fixed? | 18:07 |
| === IslandUsurper is now known as IslandUsurperAFK | ||
| === deryck is now known as deryck[lunch] | ||
| RenatoSilva | verterok: http://socghop.appspot.com/document/show/gsoc_program/google/gsoc2010/faqs#timeline | 18:44 |
| === deryck[lunch] is now known as deryck | ||
| === radoe_ is now known as radoe | ||
| === salgado is now known as salgado-afk | ||
| jam | james_w: /wave | 20:59 |
| james_w | hi jam | 21:00 |
| jam | james_w: I'm happy to add an enable/disable flag for the lp cache, what would you like the default behavior to be | 21:00 |
| jam | ? | 21:00 |
| james_w | I'm happy for it to be disable the cache | 21:00 |
| james_w | my one concern is that other people may try two in parallel and have it fall over | 21:01 |
| james_w | with a confusing error | 21:01 |
| james_w | but I don't see a way to avoid that really | 21:01 |
| lifeless | james_w: why would it fall over? | 21:02 |
| james_w | stupid stupid cache implementation | 21:03 |
| jam | lifeless: https://bugs.edge.launchpad.net/launchpadlib/+bug/459418 | 21:03 |
| ubottu | Ubuntu bug 459418 in launchpadlib "Cache is broken with multiple processes" [Undecided,New] | 21:03 |
| jam | the cache is not multi-writer safe, aiui | 21:03 |
| lifeless | ah | 21:03 |
| lifeless | closely related to 'stupidly needing a cache' | 21:03 |
| * lifeless tries for bitter | 21:04 | |
| james_w | lifeless: how would you avoid one? | 21:04 |
| lifeless | james_w: for lp apis? | 21:05 |
| james_w | yeah | 21:05 |
| lifeless | design it differently | 21:05 |
| lifeless | the amazon EC2 -style- of API's don't need a cache | 21:05 |
| lifeless | and operate hugely more efficiently | 21:06 |
| james_w | ah | 21:06 |
| james_w | ok | 21:06 |
| lifeless | specific changes are - out of band definition of APIs; a-priori defined calling method; message signing rather than cookie-auth | 21:06 |
| james_w | is it possible with bzr's Options to have no -- form? | 21:08 |
| james_w | I only need the short form | 21:08 |
| lifeless | yes | 21:08 |
| lifeless | its optparse | 21:08 |
| lifeless | (not that oauth is strictly cookie based) | 21:09 |
| james_w | well, it's not possible with Option directly as far as I can see | 21:11 |
| james_w | I can subclass | 21:12 |
| lifeless | james_w: oh interesting, its less optparse than I remember | 21:16 |
| lifeless | james_w: so no, it doesn't look like you can | 21:16 |
| jam | james_w: yeah, you'd have to override Option.add_option so that it doesn't always add '--%s' % self.name | 21:16 |
| jam | call it ShortOption, I guess :) | 21:16 |
| jam | though I'm curious why it would be a problem to have a long option, even if you don't strictly need it | 21:17 |
| james_w | just overhead | 21:17 |
| james_w | I'm taking a bunch of arguments that a subprocess takes just so I can pass them through | 21:17 |
| james_w | and it doesn't have long versions for some things | 21:18 |
| lifeless | this is for -uc and the like ? | 21:19 |
| james_w | yeah | 21:20 |
| james_w | I'm not happy with any alternative I've come with so far | 21:20 |
| james_w | and I have to override -v for verbose somehow | 21:20 |
| lifeless | you might want to override run_argv instead | 21:21 |
| james_w | oh, but I've never actually accepted verbose, so there won't be a behaviour change | 21:23 |
| jam | james_w: I've pushed an update that adds an '--lp-cache' parameter, which lets you set the lp cache directory (defaulting to None) | 21:23 |
| jam | It also includes my other small import-scripts tweaks | 21:23 |
| james_w | your happy that everyone will have to add that option if they want to make use of it in testing? | 21:25 |
| jam | james_w: well, lp doesn't provide things for you, and I'm not 100% sure where it should go | 21:26 |
| jam | for example, it could go in ~/.cache/... | 21:26 |
| jam | which is what bzr's launchpad plugin does | 21:26 |
| jam | you may or may not want to share it | 21:26 |
| jam | I think the # of people running it manually is probably low, vs how often it gets run in production :) | 21:27 |
| jam | :(, it takes 300 seconds to list the PackageToImport for gnome-panel | 21:27 |
| james_w | yes, but it's easy for me to change it once in production | 21:27 |
| james_w | jam: first time? | 21:28 |
| james_w | any quicker second time? | 21:28 |
| jam | this is the second time | 21:28 |
| jam | *with* the cache | 21:28 |
| james_w | ok | 21:28 |
| jam | there are 218 versions | 21:28 |
| james_w | well, I was thinking about the Debian cache actually | 21:28 |
| james_w | those files are larger than the LP wadl | 21:28 |
| jam | it seems it can be as fast as 100 s | 21:29 |
| jam | but it still took to 500s to get to "these packages are new" | 21:29 |
| jam | /me wonders if --lp-cache isn't actually causing it to be slower for some crazy reason | 21:30 |
| james_w | well, it still has all the round trips | 21:31 |
| jam | sure, but I'm wondering if it doubles the round trips | 21:31 |
| james_w | shouldn't do | 21:31 |
| jam | like "is this new enough: No", ok give it to me then | 21:31 |
| jam | I realize it is supposed to be sending the date in the first check | 21:31 |
| james_w | it's using etags | 21:32 |
| james_w | "Give me this file unless it matches <this one>, in which case just tell me that and I'll re-use what I have." | 21:33 |
| james_w | are you using --no-existing? | 21:33 |
| jam | no | 21:33 |
| james_w | then it may be opening a bunch of branches on LP too | 21:34 |
| jam | sure, but | 21:35 |
| jam | http://paste.ubuntu.com/378636/ | 21:35 |
| jam | it seems to be opening the *same* bzrdir over and over and over again | 21:35 |
| jam | and it still took 300s before it got the package listing | 21:36 |
| james_w | not sure why it seems to be repeating | 21:37 |
| james_w | however, the delay is just LP taking a while for a getPublishedSources call | 21:37 |
| james_w | which is unsurprising | 21:37 |
| thumper | how does bzr determine if a file is text or binary? | 21:38 |
| jam | seems to be from 'get_branch_parts' | 21:38 |
| jam | thumper: checks for \0 in the first 1024 chars | 21:38 |
| thumper | jam: is there a bzrlib method I can use? | 21:39 |
| jam | james_w: I think the issue is that: find_unimported_versions tries to open a branch locally | 21:39 |
| jam | and then falls back to remote | 21:39 |
| jam | copying that branch locally | 21:39 |
| jam | but if the remote branch doesn't exist | 21:39 |
| jam | then it just spins | 21:39 |
| jam | because it keeps getting None for the branch | 21:40 |
| jam | thumper: bzrlib.text_something.check_text_lines | 21:40 |
| thumper | jam: ta | 21:40 |
| jam | thumper: bzrlib.textfile.check_text_lines | 21:40 |
| james_w | jam: I don't think it will spin | 21:42 |
| james_w | oh, no, I see what you mean | 21:42 |
| james_w | yeah, that could be optimised | 21:42 |
| jam | yeah, for each version it tries again | 21:42 |
| jam | to open a branch that doesn't exist | 21:42 |
| jam | note that when it *does* exist | 21:42 |
| mkanat | So, looks like bzr doesn't understand query strings in URLs? | 21:42 |
| === IslandUsurperAFK is now known as IslandUsurper | ||
| james_w | I thought you mean "while target_branch is None:" | 21:42 |
| jam | it still tries to open the same branch over and over again | 21:42 |
| jam | mkanat: I would assume that we just ignore them | 21:43 |
| lifeless | mkanat: do you mean loggerhead? | 21:43 |
| mkanat | lifeless: No, I mean bzr. | 21:43 |
| mkanat | bzr merge https://bugzilla.mozilla.org/attachment.cgi?id=427066 -- doesn't work. | 21:43 |
| lifeless | hmm, we should preserve them | 21:43 |
| mkanat | (That's a bundle, at that URL.) | 21:43 |
| lifeless | mkanat: https://bugzilla.mozilla.org/attachment.cgi?id=427066/ | 21:43 |
| lifeless | mkanat: see the / - we add that, and moz barfs | 21:44 |
| lifeless | mkanat: there is a bug for this. | 21:44 |
| mkanat | lifeless: Ahh. | 21:44 |
| james_w | mkanat: does it escape them or drop them? | 21:44 |
| jam | james_w: no, we add a trailing slash | 21:44 |
| lifeless | or we may be dropping them; check a tcpdump to see precisely what happens. | 21:44 |
| mkanat | lifeless: The page that's getting returned seems to indicate that they're not getting through to Bugzilla at all. | 21:44 |
| lifeless | for opening bundles etc this is bad. | 21:44 |
| lifeless | mkanat: tcpdump time | 21:44 |
| jam | lifeless: get_transport(http+query).base has no query param | 21:47 |
| lifeless | ok, thats a bug | 21:47 |
| lifeless | we've no reason to strip query params *until* we go up a dir | 21:47 |
| jam | lifeless: ConnectedTransport._split_url calls urlutils.parse_url(url) | 21:49 |
| jam | which returns a 'path' that does not have query parameters | 21:49 |
| lifeless | sure | 21:49 |
| lifeless | just saying, that this is a bug IMO | 21:49 |
| jam | sure | 21:50 |
| jam | parse_url gets 'query' and just discards it | 21:50 |
| jam | it also discards "params" | 21:50 |
| jam | which may be a problem for colocated branches | 21:50 |
| IslandUsurper | is there a configuration option to control how qbzr highlights syntax? I want it to display files as PHP whose filenames don't end in .php | 21:52 |
| jam | lifeless: it also appears to be falling back to vfs fetch, any idea why that would be? | 21:55 |
| jam | both are in 2a format | 21:55 |
| === eugeneoden_ is now known as eugeneoden | ||
| jam | currently at 100MB downloaded according to progress bar, but only 13MB on disk... :( | 21:58 |
| lifeless | jam: on bzr+ssh ? | 22:16 |
| lifeless | jam: if you do ctrl-\ you should be able to get a backtrace; that would help determine why the fallback | 22:16 |
| lifeless | morning poolie | 22:16 |
| poolie | hi lifeless | 22:16 |
| jam | lifeless: not when running via import_package rather than bzr | 22:17 |
| jam | since it doesn't set up the signal handler | 22:17 |
| lifeless | jam: oh; then gdb python PID | 22:18 |
| lifeless | and use the gdb python macros to get a backtrace | 22:18 |
| jam | of course to do it "for real" requires waiting 5min (300s) for launchpad to finish giving me info... | 22:18 |
| poolie | jam, jeez, what are you asking it for? | 22:19 |
| lifeless | jam: you can attach live though :). still - yes, it will be painful | 22:19 |
| jam | poolie: import_package.py gnome-panel | 22:20 |
| jam | w/ ~ 218 getPublishedSources | 22:20 |
| jam | anyway, I need to be off for now | 22:20 |
| jam | I'll let it sit and spin a while | 22:20 |
| jam | gnome-control-center isn't really any faster | 22:20 |
| jam | ugh | 22:26 |
| jam | fetching "+files" redirects to "%2Bfiles" | 22:27 |
| jam | so anytime you download stuff from the librarian | 22:27 |
| jam | it involves a redirect | 22:27 |
| lifeless | or not 2b? | 22:27 |
| lifeless | isn't + 2B anyway ? | 22:27 |
| jam | lifeless: right, but "+" gets interpreted as " "I think | 22:28 |
| jam | ah wait | 22:28 |
| jam | I'm wrong | 22:28 |
| jam | +files/foo redirects to /NUMBER/foo | 22:28 |
| jam | 687.296 > GET /ubuntu/hoary/%2Bsource/gnome-panel/2.10.0-0ubuntu9/%2Bfiles/gnome-panel_2.10.0.orig.tar.gz | 22:30 |
| jam | ... | 22:30 |
| jam | < Location: https://launchpadlibrarian.net/1307193/gnome-panel_2.10.0.orig.tar.gz | 22:30 |
| lifeless | on the wire, + is 2B anyway, its in the unsafe set | 22:30 |
| jam | lifeless: yeah, I just read the redirect wrong | 22:31 |
| fullermd | Yes, I have to manually resolve those redirects every time I update the port :| | 22:32 |
| lifeless | fullermd: what, why? | 22:33 |
| fullermd | For hysterical raisins, the fetch stage of port building doesn't follow HTTP redirects. | 22:34 |
| fullermd | One reason is various server configs built by geniuses that redirect to an error page rather than returning a 404, and that sort of thing. | 22:34 |
| lifeless | is there a flag to say 'redirects are part of the spec dumbass' ? | 22:35 |
| fullermd | "Not found" is a nicer error to show a user than "hey, I downloaded this file [of HTML rather than a tarball] and the size/checksum don't match" | 22:35 |
| lifeless | fullermd: no easier to resolve when their browser shows the thing fine. | 22:35 |
| fullermd | I didn't say I entirely _agreed_ with it. It's just how it works, so either I put in the final URL, or we don't have a bzr port :p | 22:36 |
| lifeless | sure | 22:36 |
| lifeless | jsut asking if there is an override variable | 22:36 |
| lifeless | there often is in ports IME | 22:36 |
| fullermd | Potentially, by manually overriding the FETCH_CMD or something. But there's no ACCEPT_REDIRECTS sort of thing; the fetch(1) args are hardcoded. | 22:37 |
| lifeless | fullermd: sounds like a worthwhile bugfix: do it once, save time in future. | 22:37 |
| fullermd | That's more undercover reaching than I wanna do, so I just run a quick GET every update and copy in the new number on librarian. | 22:37 |
| fullermd | Yeah, it wouldn't be a huge patch to implement. Pushing through an infrastructure change like that would be more work though; I don't want it THAT much. | 22:39 |
| fullermd | Anyway, it saves LP having to service a hit every download :p | 22:40 |
| lifeless | think of the kittens | 22:40 |
| jam | james_w: as far as opening lots of branches, it doesn't help that gnome-panel is opening all of {warty,hoary,dapper,breezy,hardy,...}{,-security,-proposed,-updates,-backports,} none of which seem to exist | 22:44 |
| james_w | correct | 22:44 |
| james_w | I wasn't sure I wanted to make too many assumptions about which would exist based on others | 22:45 |
| jam | fortunately it seems to be a single BzrDir.open() probe, so a single rpc requset | 22:46 |
| jam | at least once I added a "last_branch" cache | 22:46 |
| poolie | igc, hi? | 23:01 |
| spiv | Good morning. | 23:43 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!