[12:01] SteveA: what on earth do you want me to do about this? [12:01] sabdfl: send me an email if you want ... I'll be away for 40 min .. nicole is running on zhongshan [12:01] cprov: so nicole is creating productreleases? cool [12:01] i'll keep going with my code, then we'll merge later [12:01] I'm not lifeless and I don't have the time, energy or skills to debug it - I can kill it and move your merge request out of the way if you want? [12:01] elmo: no idea. can you kill 11951 or 11878 ? [12:02] sure, I can kill 11951 [12:02] Process 11951 attached - interrupt to quit [12:02] write(1, "ALTER TABLE\n", 12 [12:02] fyi, that's the strace [12:03] this kind of thing happens when some other process is accessing that database [12:03] I wish there was a way of telling postgres to fail rather than block in these situations [12:03] killed [12:04] can you kill 11877 too? [12:05] now make can't notice it's children dieing? :P [12:07] oh, for christ's sake, every child up is hanging on writes to 1 [12:08] thanks for playing the grim reaper [12:08] I need to leave the office now. I'll talk to people about this tomorrow, try to get something sane sorted out. === kiko is now known as kiko-afk === Kinnison_ [~dsilvers@haddenham.pepperfish.net] has joined #launchpad === Kinnison_ is now known as Kinnison [12:39] Kinnison: turns out that a local test-pserver is actually easy to do [12:39] a 2 line bash script made executable, and nc, and voila [12:41] lifeless: cute [12:49] jblack: did you or someone end up helping limi out ? === BradB is now known as BradB|away [01:01] how can I get my username from python? [01:02] is os.getEnv('USER') reliable enough ? [01:02] getent stuff? [01:02] os.getlogin() ? [01:02] elmo: bingo, thanks [01:03] yay, one automagic pserver passwd file for the test suite [01:09] sabdfl: yes, now nicole is running and creating productreleases. [01:10] great work cprov [01:10] sabdfl: btw, I'm just running through main and restricted packages (1009) rather than universe, I think it's enough for test proposes [01:11] flippin 'eck, it's 00:11 again === Kinnison decides to go to bed so he can actually get up in the morning [01:11] c'y'all tomorrow === kiko-afk is now known as kiko [01:32] more pythno foo - how do I wait say a half second ? [01:38] lifeless: Sorry. Just got home. The neighbor was really sick, so I had to take care of her and her *four* kids all day. [01:38] ouch [01:38] no need to be sorry, I just needed to know if Limi was helped or not, theres no mail showing that he was. [01:38] Not by me he hasn't. Didn't know until now he had a problem. === cprov [~cprov@200.158.100.251] has left #launchpad ["Leaving"] [01:53] why dbschema.PackagePublishingStatus.PROPOSED does not exists anymore? [01:54] should I replace it for something else? [02:00] debonzi: was changed as part of Kinnison's work [02:01] best leave it as it is for the moment, and discuss with him tomorrow [02:02] sabdfl, nice, Ill comment the template to do not break the page for while. Thanks [02:03] ok [02:31] lifeless: how can i see the diff from a previous revision? === kiko is now known as kiko-afk === debonzi goes for dinner === stub [~stub@dsl-246.248.240.220.dsl.comindico.com.au] has joined #launchpad [03:01] sabdfl: in your tree? [03:01] tla changes --diffs patch-34 [03:02] lifeless: thanks [03:02] np [03:04] lifeless: [03:04] changes: illegal revision name: patch-253 [03:04] sabdfl: oh bah. [03:04] use this idiom [03:04] tla changes --diffs $(tla tree-version)--patch-253 [03:05] which is a nonsense, I'll file a bug for that. === kiko-afk is now known as kiko [03:25] anyone seen spiv around? === kiko is now known as kiko-afk === kiko-afk is now known as kiko === kiko_ [~kiko@200-206-134-238.async.com.br] has joined #launchpad === kiko_ is now known as kiko [10:29] morning all [11:14] hey === Kinnison [~dsilvers@host81-153-126-219.range81-153.btcentralplus.com] has joined #launchpad === limi [~limi@212.80-202-72.nextgentel.com] has joined #launchpad === ddaa [~ddaa@nemesis.xlii.org] has joined #launchpad [12:03] ddaa: still got that keyboard huh ? [12:04] Yes, it's cool. There is a hidden way of quitting gaim too... [12:04] Curiously, that shit only happen to me in chat clients. [12:05] lifeless: about buildbot [12:05] I assume, to have 5 successes a day, I need to run several jobs in parallel [12:05] yes [12:05] What's the catch? Use several slaves? [12:05] one slave. [12:06] its threaded. [12:06] Mh. okay... [12:06] It's probably going to complicate debugging, but for the working cases it's okay. [12:07] you could do several slaves if you want, but I'd be wary of spending a bunch of time only to find issues. (such as how to allocate the jobs appropriately) [12:07] Note that my cpu is not too fast (PIII 600MHz) so it's prolly going to be a bottleneck at some point. [12:07] sure. [12:08] go for the low hanging fruit during the day :) [12:08] Yeah. I'd like to postpone the multiple-slave issues as long as I can. [12:08] set things like xserver off when you go to bed :) [12:08] ? [12:08] it would be easier to do parallel buildbots with the same database - dual master + dual slave - for now. [12:09] well small jobs finish quickly [12:09] Good idea. [12:09] I see. I thought you were suggesting to stop X when going to be to release resources :-) === Kinnison drags lifeless over here if he wants to talk about the cscvs stuff [12:47] (unless it's public now) [12:48] lifeless: I knew I was forgetting something at the meeting. [12:49] Kinnison: cscvs is public, always has been. what we are doing with it though.. [12:49] I wanted to poke you about patchlogs which are permanently missing from rocketfuel versions. [12:49] ddaa: arh [12:49] lifeless: I was under the impression the new stuff still wasn't public [12:49] Kinnison: the code isn't yet no. I have the ok to release it, in step with our archives. [12:50] For example the reverted patch from launchpad, or the pre-tag history of soyuz. [12:50] anyhoo, dod you see the bug ? [12:50] ddaa: the pre-tag history of soyuz isn't missing, it was merged in when soyuz was merged in. [12:50] lifeless: Unlikely unless you pointed it to me [12:50] it's annoying because it breaks "missing" and probably "replay". [12:50] ddaa: copy that patch in as a merge, fine by me. [12:51] Okay. So there is no good reason why there are patchlogs permanently missing? [12:51] Kinnison: '/foo/bar' became ':local:foo:/bar' [12:51] ddaa: which ones ? [12:51] lifeless: Eww [12:51] lifeless: wait a min [12:51] lifeless: Oopsie :-P [12:51] :] [12:51] rocketfuel@canonical.com/launchpad--devel--0--patch-422 [12:51] it now uses the cvs protocol for local repositories :) [12:52] ddaa: lifeless> ddaa: copy that patch in as a merge, fine by me. [12:52] rocketfuel@canonical.com/soyuz--devel--0--base-0..patch-29 [12:52] ddaa: you shouldn't be replaying in that branch anyway, ignore that. [12:53] Bad answer. Missing is a useful command and replay is significantly faster than update here, where for some reason it always uses the apply-delta algorithm. [12:54] ddaa: the soyuz branch is not used. [12:54] its a perfectly good answer because of that. [12:54] because missing and replay don't look at that branch. [12:54] It's used by the pqm config. [12:54] ddaa: eh? [12:55] tlash dists> cat configs/canonical.com/pqm/development | head -1 [12:55] ./pqm rocketfuel@canonical.com/soyuz--devel--0 [12:56] ok, thats interesting and all. [12:56] when I next hack on pqm, I'll update the config. [12:56] Okay. [01:16] lifeless: phwoar, big changes on cscvs i see, how's that coming? [01:17] very close to having the default branch support finished [01:17] all the db & parser changes are done, am onto the scanning of tla to recreate what the parser won't tell us. [01:18] I'm hoping to run my first test-on actual data saturday. [01:18] the ProjectProductSetup page is idling though - can you please please please push through the unsconded ones there? [01:21] lifeless: thought we switched to doing this by email? i'm way behind on email [01:22] sabdfl: you were going to finish the current page... otherwise I'm just mailing it to you again. [01:22] are you using the new project/project creation pages, and the links to setup sourcesource there? [01:22] but I can copy the current page and mail to you if you like. [01:22] sabdfl: when they go into launchpad, yes. [01:22] i'll work on it in the wiki [01:23] david is working on the research for this too now, so we'll be putting some volume up I hope. [01:23] would you like me to have him mail you & me rather than adding to the wiki page ? [01:23] hmm.. looks to me like a bunch were added after i thought we switched to email [01:23] sabdfl: nope. [01:23] onest injin [01:24] lifeless: we have a script which will pull details from freshmeat and sourceforge [01:24] celso has been working on that [01:24] sabdfl: cool. I presume we still need to determine the mapping tho. [01:25] What I'd love is to get you off the critical path - so I can go 'hey, xyzlib worked in test, goto launchpad, and just do it' [01:25] yes, what i'm really saying is that there are going to be about 300 projects / products setup automatically [01:25] that will help enormously. [01:25] we are just running some tests this week and next week, then we can put it into production [01:25] the gnome ones helped too, as as they go good we can just make them. [01:27] hmm... project db methinks should be libdb [01:27] berkely database is more than a lib [01:27] the lib is what most folk use is all. [01:28] "db" is confusing, what about berkeleydb [01:28] sure. [01:28] i don't mind the product name [01:28] is libdb1-compat part of the same project? [01:28] IIRC no. [01:29] someone wrote it because they needed it [01:29] I did check that when doing the page. [01:29] if we edit the names in the db, will it break the sync process? [01:29] or does the sync remember what it was told to use when it started? [01:30] once I put the sync gold, the name is set forever, until I go change it [01:30] the project name can vary wildy, it won't care. === limi [~limi@212.80-202-72.nextgentel.com] has joined #launchpad [01:31] spiv: sqlos patch applied [01:31] lifeless: Thanks. [01:31] Hmm, #1922 just struck again. === cprov [~cprov@200.158.100.251] has joined #launchpad [01:31] 3rd time today. [01:31] (after two days nothing) [01:31] and if you change it, does it tag into a new archive/cat-branch-version from the old one? === spiv roboots [01:32] Bug 2005 resolved: Cache prevents to use new users without restarting launchpad [01:32] https://bugzilla.warthogs.hbd.com/bugzilla/show_bug.cgi?id=2005 [01:32] sabdfl: yes, changing involves a manual tag at the moment, and change the sync in launchpad. after that the sync carries on automatically. [01:32] ok [01:32] that is partly by design, partly optimising for the common case. [01:33] the design bit is - we don't want these things randomly changing all over the place. so until we have a well articulated policy, that I can put into code, I don't want the code to have that 'feature'. [01:33] Kinnison: here's an example: [01:33] ':pserver:anonymous@127.0.0.1:/home/robertc/source/canonical/buildbot/launchpad/sourcecode/cscvs/,,repoCatalog' != ':pserver:anonymous@127.0.0.1:/srv/importdtest/buildbot/launchpad/sourcecode/cscvs/,,repoCatalog' [01:34] oh, bah. [01:34] thats just me being brain dead, nm, carry on, etc etc === Kinnison tickles lifeless === limi [~limi@212.80-202-72.nextgentel.com] has joined #launchpad [01:40] Kinnison: this is it (its still happening...) [01:40] print CVS.protocol.CVSRoot('/srv/importdtest/botslave/buildbot-jobs/autoconf-HEAD-import.job/autoconf@arch.ubuntu.com/autoconf--MAIN--0/cvs_temp_repo') [01:40] :local:/srv/importdtest/botslave/buildbot-jobs/autoconf-HEAD-import.job/autoconf@arch.ubuntu.com:/autoconf--MAIN--0/cvs_temp_repo [01:40] I haven't quite fixed it :[ [01:40] lifeless: Aah it's not liking the @ inside the path [01:40] the earlier bug was allowing / in the hostname [01:40] lifeless: Dunno how you'll fix that except to disallow / inside users and passwords in the regexp [01:41] Kinnison: probably break it into ()|() [01:41] possibly [01:41] I'm not sure that the @ is the problem yet [01:42] it smells like it to me [01:42] ok, now I'm sure. [01:42] z.username -> [01:42] /srv/importdtest/botslave/buildbot-jobs/autoconf-HEAD-import.job/autoconf [01:42] yeppers [01:44] thankfully, your test suite is comprehensive enough to catch me being an idiot regression wise :) === Kinnison grins [01:45] username can't have / can it ? [01:45] I sincerely hope not [01:45] that fixes it trivially. [01:46] ^:([^:] +):(?:(?:([^:/] +)(?::([^:/] +))?@)?([^:/] +)(?::(\\d+))?:?)?(/.*)$ [01:46] method username password hostname port path [01:46] yep [01:47] possibly wrong to put the / in the password bit, but hey, cross that bridge when the bug report lands [01:47] aye [01:47] this should finally allow me to not get booted off the kde anonymous server. [01:47] so we can do kdelibs as soon as spiv makes me proud [01:48] hurrah [01:48] (fingers crossed etc etc etc) [01:48] lifeless: :) [01:48] have I mentioned how great it is to have this native facility ? === Kinnison grins. You're welcome [01:50] you might find looking at the current protocol / Protocol / pipes interesting [01:51] you had a design flaw - using a list of handles isn't sufficient to do cleanups properly. [01:52] oh? [01:52] lifeless: ok, project is xiph, product is paranoia, productseries is Paranoia-III [01:52] productseries ? [01:52] yes, name for a set of releases === ddaa gets a tarfile.ReadError and goes to find the bugzilla [01:52] like "head" [01:52] ah. [01:52] or "2.0" [01:52] ok, thats not in production yet. [01:52] or "Paranoia-III" [01:52] lifeless: no, it's last night's work, though the db stuff has been there for a while [01:53] ok [01:53] creating a productrelease will require specifying a series for that release shortly [01:53] this allows us to do better magic when we have intermingled releases of, say: [01:53] apache 1.3 [01:53] no, better example would be: [01:53] gimp1.3 [01:53] gimp2.0 and [01:53] gimp-cvs [01:54] (2.1) [01:54] lifeless: where do the buildbot bugs live? [01:54] create three series, and for each release just tell it which series it belongs to [01:54] ddaa: bugzilla.canonical.com or bugzilla.warthogs.hbd.com - the original main bugzilla [01:54] sabdfl: so a series is roughly a branch :) [01:54] roughly, yes [01:54] I was redirected to bugzilla.no-name-yet.com... [01:54] ddaa: yeah. let me give you a direct url [01:55] https://bugzilla.warthogs.hbd.com/bugzilla/show_bug.cgi?id=2005 [01:55] thats in the right system. [01:55] I think we may need to take out the python tar module, it appears to be slightly crackful. [01:57] thanks [01:59] lifeless: i think keybuk uses that heavily for sourcerer [01:59] sabdfl: he was saying it had really bad problems [01:59] which I filed to look into later [01:59] k [01:59] so, the public name of buildbot is Hoover, right? [02:00] yes [02:00] out install of buildbot sucks. [02:00] s/out/our/ [02:00] :) [02:00] cprov: around? [02:03] Kinnison: got a few minutes ? [02:03] got a change in behaviour from cvs external rlog to the native implementation. [02:04] https://chinstrap.warthogs.hbd.com/importd/devel-status/libbonobo-HEAD-import.job/events/289/log (if you have time) [02:06] sabdfl: yes [02:06] sabdfl: I'm testing the new dump including productreleases [02:06] Just to be sure I grok bugzilla right, there are currently 4 bugs for Hoover whose status is anything but "closed"? [02:06] https://chinstrap.warthogs.hbd.com/importd/devel-status/kdelibs-HEAD-import.job/events/273/log for folk that like to give browsers a hernia [02:06] stub: what's the correct way of doing something like this: [02:06] * SELECT pg_catalog.setval('schema_id_seq', pg_catalog.select(SELECT max(id) FROM Schema"), true); [02:06] ddaa: sounds about right. [02:07] cprov: could you point me at that too please? want to see how it looks [02:07] lifeless: is this a hint that I should enforce mod_deflate? :P [02:07] cprov: do you use the ProductRelease object from canonical.launchpad.database or a custom Soyuz one? [02:07] elmo: wink wink nudge nudge [02:07] sabdfl: Dunno - what are you trying to do? [02:08] elmo: I think that would actually make browsers less happy, not more. [02:08] i was looking to merge your code now if it's still Soyuz* [02:08] Looks like you are trying to reset the primary key sequence to the next highest value? [02:08] stub: trying to be able to add rows of sampledata then reset the serial counter to the max current [02:08] correct [02:08] sabdfl: mailed you a high-priority mapping - kdelibs. [02:08] sabdfl: I'm using custom one, since we aren't use SQLO in those scripts [02:09] cprov: oh, ok [02:09] that's probably fine then [02:09] why not use SQLO? [02:09] DISTINCT etc? [02:09] lifeless: that's just a big ass file, right? [02:09] sane queries etc? [02:09] sabdfl: in zhongshan:/home/cprov/ lp_dump-productreleases-mr.sql [02:09] random failures etc? [02:09] lifeless: buildd.d.o has those too and .bz2's them by default - the browser seems to cope okay [02:09] right, those reasons. [02:10] elmo: dynamic, scrolls as you wait [02:10] sabdfl: Not sure - I've only used alter sequence which won't let you embed a SELECT in it [02:11] stub: no worries, just thought i'd play "throgh crackful queries at the BDBA and see how he handles" today :-) [02:12] sabdfl: yes, many simple feature are harder to do using SQLO, maybe my fault, but using simple "pg" the things shows up faster, same strategy used by elmo and Kinnison [02:12] works for me [02:12] sabdfl: select setval('bug_id_seq', (select max(id)+1 from bug)); [02:13] let's help spiv and bradb by telling them wat tweaks would make sqlo more useful in scripts though [02:13] stub: nice, thanks :-) [02:14] sabdfl: ERROR: column "productseries" does not exist SELECT id FROM ProductRelease WHERE productseries = 466 ? [02:14] elmo: should i be able to scp off zhongshan using my normal ssh keys? [02:14] ok, I'm going to crash, started at 7am :| [02:14] spiv: mail me anything you'd like me to investigate. [02:15] ddaa: good luck, please mail me for info etc. [02:15] cprov: there's an sql patch which went in today for that [02:15] I'm seeing RMS speak tomorrow, so will not be around all that much - will be working offline mostly. [02:15] it was in the pending queue for a day or two [02:15] night all. [02:15] night robert [02:15] lifeless: you have hint about the tarfile error? [02:15] i seem to get that one for everything [02:15] ddaa: oh. [02:15] what ones specifically ? [02:16] if its the X.org ones, its cause that module can't hack the 800MB tarball. [02:16] so far: zenity, yelp, xrestop [02:16] (or, that the URL for the tarball is wrong) [02:16] hmm, dunno offhand. pdb is your friend I think [02:17] lifeless: what i was planning to ho [02:17] run the slave with twistd -nf buildbot.tap [02:17] and throw an import pdb;pdb.set_trace() at the appropriate place. [02:17] ack, I had put that one into a sticky note a few days ago :-0 [02:17] ok, I'm out... [02:17] great. [02:17] night! [02:17] sabdfl: yep, I see, we have now productseries... uhmm it crashes my dump entirely [02:17] sweet dreams [02:18] sabdfl: if you have the .ssh/config set up right, yes [02:18] what's the voodoo for that? [02:18] Host *.ubuntu.com [02:18] ProxyCommand ssh mark@chinstrap.warthogs.hbd.com nc -q0 %h %p [02:18] add that to the other stuff [02:18] for the zhongshan host only? [02:19] do it for *.ubuntu.com, then you get the other ubuntu.com hosts for free as we migrate [02:20] sabdfl: anyway, using that I pointed you, you can see productseries [02:20] can I have separate entries for *.ubuntu.com [02:20] and zhongshan [02:20] ? [02:21] cprov: so your code created productseries entries, and productrelease entries? [02:21] but didn't know about the new field to tell a productrelease what productseries it is part of? [02:23] sabdfl: err, sure if you want, can't imagine why you'd want to tho? [02:24] sabdfl: yes, I missed the field productseries in table Productrelease, so it is almost unuseful :(, sorry [02:32] spiv: is there some magic to prevent twistd from decorating the output of the buildbot? The "2004/10/14 14:31 CEST [Broker,client] " prefix breaks the integration of emacs with pdb. === SteveA [~steve@office.pov.lt] has joined #launchpad [02:35] elmo: wow that was exciting [02:35] just DOS'd myself with a bad .ssh/config [02:38] want to try this at home? [02:38] Host zhongshan [02:38] compression yes [02:38] HostName zhongshan.ubuntu.com [02:38] Host *.ubuntu.com [02:38] ProxyCommand ssh mark@chinstrap.ubuntu.com nc -q0 %h %p [02:38] elmo: was hoping to be able to put the proxycommand in a *.ubuntu.com entry [02:38] and then use separate entries to map host to ubuntu.com [02:40] Kinnison: what do I need from librarian to run gina now ? [02:43] sabdfl: turn on tab completion [02:43] cprov: You need a running librarian [02:44] then it's just zhon, rather than zhonghan or zhongshan.ubuntu.com [02:44] elmo: still not working, even if I specify things completely === limi benchmarks Javascript [02:44] Host *.ubuntu.com [02:44] compression yes [02:44] ProxyCommand ssh mark@chinstrap.ubuntu.com nc -q0 %h %p [02:44] sabdfl: chinstrap.warthogs.hbd.com dude [02:44] sabdfl: yes, I know the half and half things sucks, sorry [02:44] sabdfl: 20K entries is what we are expecting for packages, roughly? [02:45] Kinnison: do you have it zhongshan, in a easy way that I can copy? [02:45] You edit lib/canonical/librarian/server.tac to set a path to the storage and to choose a port for it [02:46] then you run (from the level where you'd make run for launchpad) twistd -noy lib/canonical/librarian/server.tac [02:46] If you don't have it already; then prefix that twistd command with 'PYTHONPATH=lib ' [02:47] Kinnison: ok [02:48] cprov: It's important to remember that the database and filesystem path combination work together for the librarian [02:48] so if you blow the db away; remember to blow the filesystem it created away [02:48] Also if you copy the DB, copy the filestore too [02:48] limi: for a single distro that would be the max for now, yes [02:48] ok [02:48] cprov: *or* mod gina to make the librarian interface optional :-) [02:49] Kinnison: I thinking in this possibility now :) === cprov has just give it up when looks the code [02:54] ew [02:54] the tarfile error is caused by getting a 401 from chinstrap... [02:54] (auth required) [02:55] in zenity.info: [02:55] CvsTarFile: http://chinstrap.warthogs.hbd.com/~jdub/cvsballs/zenity.tar.bz2 [02:57] So... what would be the magic trick to make that work from my workstation? [02:57] elmo: any idea? === SteveA [~steve@office.pov.lt] has joined #launchpad [03:01] cprov: eh? === ddaa investigate having pycurl handling ssl and auth === ddaa wishes his fingers knew how to write english. [03:04] Kinnison: yep, It will take so long to mod gina to work w/o librarian and I'll take the risc of kill you work, I've just perform some surgery on my DB and recover the previous generated gina data [03:05] Kinnison: but I'll be glad if you can do the modularization of gina :) [03:05] cprov: Maybe when I've sorted this poolbuilder === Kinnison is currently eating his lunch; then I'll get back to writing lib.canonical.lucille.publishing.SourcePackagePublisher [03:07] Kinnison: thanks, enjoy you lunch ! === BradB|away is now known as BradB === limi finds out what he suspected: javascript isn't the fastest language to do string comparisons in [03:09] :) [03:26] sabdfl: so, now it is ok, nicole is generating correct productreleases and I modified it in the way that you can take partial dumps when it still running to have a kind of DB snapshot [03:27] sabdfl: I have one im my home with ~10 projects ...if you want you can do you own now (~30) === ddaa wonders if curl could use .netrc of http authentification... === debonzi [~debonzi@200.158.100.251] has joined #launchpad [03:44] Looks like it works... [04:03] cprov: could you do a dump for me please, and point me at it? [04:04] sabdfl: of course, just a second [04:05] limi: figured out how to make javascript fly? [04:05] sabdfl: scp zhongshan.ubuntu.com:/home/cprov/lp_dump-productrelease-partial.sql.gz [04:07] cprov: thanks! [04:07] do I just drop my db and then use this to recreate it? [04:07] sabdfl: "not fly", as the case may be - the conclusion is that we should use XML-RPC with an inline search widget to do this server-side - with a fallback on the button to do a traditional search if you don't have JS [04:07] dropdb foo; createdb -E UNICODE foo; zcat bar.sql.gz | psql foo [04:07] Javascript just won't handle 20K items (tested with 5K) [04:08] limi: absolutely no way to do this client side? [04:09] nope, too many entries - but it is client-side in the sense that you don't leave the form [04:10] stub: Can we drop the dependence on plpython? The security concerns it adds don't seem worth it for 10 lines of Python validation code. [04:10] sabdfl: dropdb XXX; createdb -E UNICODE XXX, psql -f XXX [04:10] What is the security concern? [04:11] The fact that pqm has to run as a superuser on the DB for us to checkin changes ;) [04:12] I don't think that actually works yet, but it was supposed to yesterday, until we hit that roadblock. [04:12] a postgres superuser, although I'm discussing this with steve on habber atm. Why is that a problem? [04:12] stub: It's a big risk to take if the production database is on the same machine as a non-production one. [04:13] And, well, it's 10 lines of Python. :) [04:13] I wasn't aware we had that situation anywhere. [04:14] I'm sure elmo could provide a more detailed explanation of the risks involved though. [04:14] It will be more one day. It is trivial to redo it in pl/pgsql so I'm not fussed about the current code but it is more for future expansion. [04:15] It means that pqm has write access as the postgres user, so they can trash all the postgres databases on that server if they want. [04:15] yep, that's the most obvious risk. there are probably other things though too. [04:16]