IRCloggy #git 2006-07-28

Logs Search ←Prev date Next date→ Channels Documentation

Provider of IRC logs since 2005.
WARNING: As Freenode became unjoinable and lost all warnings in topics, we cannot log channels on Freenode anymore.

2006-07-28

pasky telmich: where can I pull from?00:05
telmich: ah, now I see :)00:05
CIA-10 Cogito: pasky * r45b96b2706c2 /cg-tag-ls: Merge with http://unix.schottelius.org/git/cogito-original.git/00:06
→ tchan joined00:41
dwmw2_gone → dwmw2_tw00:51
CIA-10 Cogito: pasky gitgit-stash * r0107a1ac15eb / (cg-Xfetchprogress cg-fetch): Support for git-*-fetch --stdin - massive fetching speedup00:54
pasky (gitgit-stash is my local branch I just created which depends on unreleased or even unmerged-yet Git features; I probably won't publish it unless someone asks for it specifically, and I don't even know if anything significant will end up in it)00:55
cw666 aside from the fact you can argue this is a retarded thing to deal with, one-file-per tag/referernce really blows if you have 10s of thousands of these00:56
has anyone considered maybe moving the refs into their own file which is access using a hash or something?00:57
→ goncha joined00:57
pasky Junio, when talking together at OLS ;-)01:03
but I suppose more as a sorta might-be-nice-to-have-or-perhaps-sometimes-in-the-future-but-i-would-take-a-patch-if-someone-submitted-one01:04
dkowis thems some hyphens01:05
CIA-10 Cogito: pasky * r647de0de67c9 /TODO: TODO: Unify hooks with Git01:06
cw666 pasky: was that in response to me or someone else?01:29
antarus needs to figure out git-fetch too01:29
pasky cw666: to you01:32
cw666 pasky: what did junio think might be a good idea? the problem with a maga-file replacing this is that you need to watch it does get corrupted/hosed01:32
especially when updated01:32
CIA-10 Cogito: pasky * r855390731c09 /cg-patch: Fix applying of files moved to a new directory01:41
pasky cw666: the idea was something like a berkeleydb, and yes, corruption issues was exactly what I pointed out01:49
IIRC we just sorta nodded to that objection and moved on ;)01:49
you just need to be careful and choose a db well-resistant to that, it should be possible01:49
or we might do something entirely custom, at worst01:49
cw666 pasky: bezerkelydb can be made transactional and safe if carefully used01:51
alp pasky: i see you git patches, interesting. typofix for one of them though, 'to except' has pretty much the oppisite meaning of 'to accept' which you meant01:51
cw666 though clearly most people don't do that because well, it breaks often in many cases01:51
alp er, opposite01:51
antarus hrms01:52
antarus I think this will not work out well at all01:52
pasky alp: when?01:53
oh, read "see" as "sent"01:54
what a pity ;)01:54
alp heh, it's a universal law that every attempt you make at fixing typos involves more of them, my sentence was horrible ;-)01:55
→ tko joined01:56
pasky thanks01:57
CIA-10 Cogito: pasky * r5966d8c3c0f1 /TODO: Removed the school assignment notes01:59
alp pasky: so this makes it possible to update a bunch of totally unrelated repos all in one go?02:00
pasky nope, sorry02:03
antarus hrm, the kernel uses git, I wonder if their stats are as bad as mine...02:04
pasky stats?02:04
antarus I should be able to fetch the kernel sources and then use git-clone to test..02:04
pasky: git-clone uses too much ram, is 3x slower than cvs02:04
it seems like the ram thing is abnormal though..02:04
bartman but cvs doesn't retain history on the client02:04
it's not a fair comparison02:05
antarus the speed I'm not worried about02:05
bartman then you should use rsync02:05
pasky antarus: what protocol are you cloning over?02:05
antarus ssh02:05
pasky and what git version?02:05
antarus 1.2.402:05
pasky ow02:05
antarus old?02:05
pasky I'm 90% sure there were related fixes happenning later than that02:06
so the latest git might perform noticeably better.. might be worth a try02:06
antarus I'll poke ramareth02:06
pasky alp: what protocol are xorg people fetching over?02:06
antarus let me fire up rsyncd on a alternate port and see how it performs02:07
pasky antarus: rsync is severaly discouraged for productional use, so I don't think it's worth the time02:08
testing the latest git version time would be much more time-effective, imho02:08
→ krh joined02:08
alp pasky: git-daemon for pulls, ssh for pushes i believe, since ssh for pulls even by the develpers was overloading the server at some point02:09
pasky hmm02:10
git or cogito?02:10
(if there's some general line)02:10
alp every xorg/freedesktop person i've spoken to seems to use core git, but nobody has anything against cogito and i guess people would be happy to use a cogito tool if it gave this benefit02:11
pasky: apart from the developers, the other half of the noise about git being slow is coming from distribution packagers and power users who remeber the days of cvsup02:12
so, that's the read-only, or small patch producing crowd02:12
pasky will you be around for a while yet?02:13
I might be on to something02:13
alp yeah, will be around for a couple of hours02:13
pasky cool02:13
antarus pasky: git-daemon still uses tons of ram, 1gb+02:13
but our repo is 1.4gb02:13
alp: any idea how big your biggest repo is?02:13
pasky antarus: it'd be best if you could drop a mail at the mailing list02:14
someone is sure to look into it02:14
antarus let me do that now02:14
else I'll keep putting it off02:14
pasky thanks :)02:14
antarus I'm sure you enjoy me bringing it up here every couple of days ;)02:15
alp antarus: all small scale, linux is the daddy here by far02:16
antarus nods02:16
antarus I guess I'll grab a copy of linux while I'm waiting and compare ;)02:16
→ agile joined02:19
antarus yikes02:30
kernel is tiny02:30
alp antarus: maybe there's something wrong with your history? the linux kernel repo is massively active and has a few years data02:39
antarus: it could be that redundant commits got in during the translation, or maybe auto-generated data is being versioned?02:39
DrNick complete history, including the graft, is ~310 MB02:39
antarus alp: we have a bunch of nasty "commit the entire repo" commits02:41
I doubt they help02:41
alp you might want to step through the import process and modify or eliminate the useless commits, just like using a good debugger02:42
antarus mm02:42
the import process was 15 days02:43
alp well, if you can figure out what the silly commits are, you can just script a filter that will step over them02:43
antarus nods02:43
alp i hand-built a repo for a tiny repository of mine from tarballs, and fixed the random permissions and directory layout at each revision so i had a sensible history. even though it didn't match reality, in my particular case this wasn't a problem because the code was unreleased02:44
antarus nods02:45
antarus pasky: you win, it appears 1.4.1 is doing a ton better03:14
→ boto joined03:16
pasky cool03:24
hmm03:34
Google launches SF.net-like code hosting tool03:34
and they just have subversion03:34
pah03:34
$ git-fetch-pack --repo=/home/xpasky/git-pb/.git master o/pu --repo=/home/xpasky/cogito/.git v0.17 telmich03:37
Segmentation fault (core dumped)03:37
ok, I s'pose it still needs some fine-tuning03:37
→ xjjk joined03:51
→ Tv joined03:56
→ c joined04:13
pasky alp: I almost have it04:27
alp pasky: am right here, good luck04:31
pasky d035cf5f74307963c62b71137710390880857ad4 refs/heads/o/pu /home/xpasky/git-pb/.git04:47
807c974d3c8427c47f601d0167e3cf59b33c76cc refs/heads/v0.17 /home/xpasky/git-devel/.git04:47
ok, I guess it works :)04:47
now a simple wrapper for that04:47
→ lilo joined05:12
telmich pasky: yes, currently only minor changes, but I had some questions regarding cogito05:19
pasky: cg-branch-ls: I would not call 'mkdir -p' each time, but remove it and do only '[ ! -d ... ]'05:20
pasky: so creating actually takes place in the add-part05:21
uih, gitweb is broken on kernel org05:23
http://www.kernel.org/git/?p=cogito/cogito.git;a=commit;h=fb727fe11b6f6d680a4f048f4d28db6a45f32bf405:23
→ kanru joined05:24
telmich http://www.kernel.org/git/?p=cogito/cogito.git;a=blob;h=385addc532d6232d690e745743ed8a2664270503;hb=c3330805bd3aa0d60a1ff1f7a38e1adac2c40f63;f=cg-Xlib05:24
...05:24
pasky: are you the one, one should tell that?05:24
pasky what's wrong there?05:24
telmich well, select any project05:25
in the summary, the partly shortlog is missing05:25
if you select a commit in shortlog05:25
it says: 403 Forbidden - Unknown commit object.05:25
pasky seems to look fine in cogito...?05:25
ah05:25
telmich no, not fine in cogito05:25
pasky that's likely some usual rsync desync05:25
telmich I just tried cogito/cogito.git and git/git.git05:25
pasky hmm?05:25
well cogito summary looks fine here05:26
telmich ahh05:26
hrm, no the standard project listing at http://www.kernel.org/git/ also vanished05:26
do you have access to that machine / can you check whether permissions are simply wrong now?05:27
ah, now it works05:27
no, not really05:28
strange, I hit reload, it works, I hit reload I get the 40305:28
well, got to drive to work, just wanted to mention it05:28
pasky probably the two kernel.org mirrors are desynced05:33
alp: check your mbox05:46
alp pasky: looking05:47
will patch up and give it a test run now05:48
pasky cool05:49
and I'm gonna get some sleep :)05:49
alp thanks for looking into this pasky, 'night05:54
→ nwf joined06:57
dwmw2_tw → dwmw2_KHH06:59
telmich pasky: how can/should I in general notify you of changes in cogito when I do them?07:02
→ robfitz joined07:03
telmich pasky: perhaps droping an e-mail when doing cg-push using some hook?07:04
pasky: and is there a hook-listing for cogito available?07:14
→ kanru joined07:36
→ GyrosGeier joined07:50
→ ferdy joined08:39
→ vegard joined09:13
→ benlau joined10:36
robin How to I undo changes on a single file?10:48
neuralis robin: git checkout <file>10:55
→ nwf joined11:17
→ Temaris joined11:25
robin neuralis: ah, that simple :) thanks11:28
← makx left11:45
→ benlau joined11:57
dwmw2_KHH → dwmw2_HKG12:16
→ kanru joined12:25
→ boto joined12:50
pasky alp: so, any results already? ;)13:04
telmich: no hooks list I'm afraid, but you can grep HOOKS cg-*13:04
telmich: for notifying about changes in repo, git-request-pull might be attractive13:07
telmich pasky: so which hook would be called at cg-push?13:08
ah, update and post-update13:09
sorry for asking, was just confused by grep HOOKS, because they are written lower case13:09
pasky: do you have an e-mail for me, so I can notify you of changes?13:10
→ agile joined13:21
→ coywolf joined13:26
→ _mkydiver joined13:27
Elrond Just to ask the standard FAQ: Is gitweb on kernel.org for git.git b0rked?13:34
moh I've seen people say that gitweb on kernel.org might be borked right now13:37
Elrond Ahh. So it's not me. Reloading sometimes helps.13:38
moh give it some time. Its sill pretty early on the west coast :-)13:38
Elrond Ahh, yeah.13:39
→ chris2 joined13:51
pasky telmich: [email@hidden.address] but I wouldn't be afraid of cc'ing [email@hidden.address] too13:57
that is, if you won't push too often ;)13:57
(I don't mind getting one message per push, but it's not good for the mailing list)13:57
→ krh joined13:59
telmich pasky: narf, I just saw that the hook is called on remote side13:59
pasky yes14:04
isn't that good? you can push from various places and it'll always work14:04
telmich well, actually on home.schottelius.org runs RSBAC (security framework for linux) which does not allow me to mail normally14:05
it's interesting, yes, I am not sure which method I would really prefer14:05
pasky if you need them, feel free to _add_ hooks14:06
:)14:06
telmich I am planning to create a document listing all cogito hooks, would you like it to be in asciidoc format?14:06
pasky sure14:33
dwmw2_HKG → dwmw2_gone15:00
→ xjjk joined15:29
pasky woah?15:38
do I understand it right that binary patch is not revertable?! :/15:39
that's bad15:39
→ segher joined15:49
Elrond What was the tool to find out who changed a line last?16:49
Ahh, git-whatchanged -p and / in less helped ;o)16:52
→ anholt_ joined17:54
→ nwf joined17:58
→ anholt_ joined18:03
→ matled_ joined18:03
matled_ → matled18:05
→ robin joined18:05
bartman there seems to be a clash between tracking a branche 'name' using .git/remotes/* and the remote having a .git/branches/'name'18:06
I keep getting...18:06
error: no such remote ref refs/heads/public18:06
to clarify: remote has many branches, one of them is 'public' and it has refs/heads/public in .git18:07
locally I want to track all of it's braches including 'public' using .git/remotes/origin18:07
when I git fetch I get the above error18:07
but if I locally create a .git/branches/public with the right #public URL, I can git fetch public w/o a problem18:08
... is this known? new? expected?18:08
btw, git-ls-remote for the remote shows the refs/heads/public18:09
→ boto_ joined18:18
→ alley_cat joined18:19
bartman hmm, that wasn't the whole story18:42
my .git/remotes/origin has:18:43
Pull: refs/heads/public:refs/heads/public18:43
Pull: refs/heads/public:refs/heads/origin18:43
if I remote one of them it works18:43
→ Tv joined18:43
bartman so it looks like I cannot update two refs locally based on a remote ref18:45
lesson learnt18:45
→ Drakeson joined18:46
← Drakeson left18:46
→ xjjk joined19:38
→ nwf joined20:13
→ _JusSx_ joined20:18
cw666 is there any way to have a RO branch and a RW branch exist in the same repo?20:19
filesystem permissions clearly don't work in this case20:19
robin tried the sticky bit?20:21
cw666 erm, how is that going to help?20:26
i mean, it will retrict who can update the heads i guess20:26
but to needs to be a group of lusers where i dont have a gid set aside and usable20:27
(nss is ldap based from some AD crap, making groups is a problem)20:27
bartman cw666: try to play with permissions of .git/refs/heads/BRANCH_NAME20:29
cw666: see if a user that does not have write permissions to that file can push20:30
cw666 bartman: yeah, tat20:30
grr20:30
stupid ' key :-)20:30
bartman: yeah, ill do that for now20:30
bartman the worst that could happen (I think ... and please don't quote me on it) is that git will create the object files, but no one will be able to reference them20:31
robin cw666: just a guess20:34
it seems git writes a /new/ file when it updates the head20:35
having a / in the branch name may help20:36
cw666 yeah, im not that bothered, there are limited people who should be able to break things so ill probably just resort to a LART if (when) that happens20:36
bartman cw666: you can probably do the test in the commit hook20:37
although that may not work if they are using ssh://20:38
cw666 bartman: except this isn't a commit, it's when someone pushed tags they can break it20:38
the commit hook asumes they won't frab that locally surely?20:38
cw666 fights with git-cvsimport21:01
bartman just be happy it's not cvs-gitimport :)21:05
cw666 git-svnimport rocks, for some reason though cvsps is giving me pain21:09
telmich pasky: pull-request, mails do not work on home yet21:15
→ DrNick joined21:20
→ robin joined21:23
→ normalperson joined22:39
→ beu joined22:50

Logs Search ←Prev date Next date→ Channels Documentation