| 2006-10-27 |
|
DrNick
| does this gittorrent thing trust all peers? | 00:05 |
|
mugwump
| it trusts them, sure. | 00:08 |
|
| about as far as it can throw them. | 00:08 |
|
DrNick
| so, given that I can generate arbitrary files for a specifica SHA1 hash, why is this a good idea? | 00:08 |
| → cworth joined | 00:09 |
| → GeertB joined | 00:17 |
| → Newsome joined | 00:30 |
| → spearce joined | 00:34 |
|
mugwump
| DrNick: you can? cool... care to show me? :) | 00:36 |
|
| There's no reason that the next version of the protocol couldn't use SHA256+ for references | 00:38 |
|
| Other than a bit of wasted time re-computing object IDs using a slower digest. But I think that's a version N thing, where N > 1 | 00:38 |
|
| spearce catches up on the IRC logs... | 00:44 |
|
spearce
| GitTorrent - very interesting idea. | 00:44 |
|
| pasky tries to remember how the hell did he got from fixing up random repo.or.cz bugs to designing a mechanical turing machine on a nearby envelope | 00:45 |
|
pasky
| git is absolutely horrible :/ | 00:45 |
|
| pushing a git instance to localhost took a fair number of minutes | 00:45 |
|
| of course because of the stupid git-receive-pack idea that it has to unpack all the objects | 00:46 |
|
spearce
| Nico "fixed" that recently. :-) | 00:46 |
|
pasky
| this makes pushing any kind of non-tiny project completely impractical since this will likely take hours | 00:46 |
|
| is his fix in next already? | 00:46 |
|
| and how did he "fix" it? :) | 00:46 |
|
spearce
| its in pu right now and he fixed it by completing the "thin" pack to be whole and then keeping that pack as a pack. | 00:47 |
|
| but there's apparently a race condition now with git-repack -a -d. :-( | 00:47 |
|
pasky
| hmm, how? | 00:47 |
|
spearce
| he appends the missing bases onto the end of the thin pack, then writes that out as a pack with its associated index. | 00:47 |
|
pasky
| does completing the thin pack involve pulling all the history in it as well? | 00:48 |
|
| or just trees/blobs? | 00:48 |
|
spearce
| any missing delta base, so any type of object could get appended onto the end. | 00:48 |
|
pasky
| ah but only along delta chains, yes, I'm stupid | 00:48 |
|
spearce
| your working on envelopes. its ok. :) | 00:48 |
|
pasky
| yeah I should go back to pipes and lead balls | 00:49 |
|
ShadeHawk
| Has anyone have benchmark comparing git network performance with other SCM up there? Like Mercurial wire protocol? | 00:54 |
| → anholt joined | 01:02 |
|
mugwump
| ShadeHawk: I compared 'svk sync' with 'git clone' on my use.perl.org journal | 01:17 |
|
| 13,000s to fetch 11,000 revisions vs. ~70s to fetch 6,000 | 01:18 |
|
spearce
| were they the same revisions? | 01:20 |
| → dst_ joined | 01:43 |
| → beu joined | 01:45 |
| → Gitzilla joined | 01:51 |
| → beu joined | 02:00 |
| → robfitz joined | 02:06 |
| → xjjk joined | 03:05 |
| → anholt joined | 03:44 |
| → spearce joined | 04:11 |
| → lyakh joined | 05:11 |
| → spearce joined | 06:30 |
| → normalpe1son joined | 07:39 |
| → ferdy joined | 07:44 |
| → segher joined | 08:50 |
| → chris2 joined | 09:18 |
| → normalperson joined | 09:23 |
| → robfitz joined | 10:22 |
| → chris2 joined | 10:37 |
| kittah → dkowis | 11:31 |
| → boto joined | 11:54 |
| → cworth joined | 12:05 |
| → Whoopie joined | 12:19 |
|
Whoopie
| pasky: ping | 12:20 |
| → timlarson_ joined | 12:46 |
| → luks joined | 13:03 |
|
kampasky
| pong | 13:04 |
|
| Whoopie: missed you by exactly 10 seconds yesterday ;) | 13:05 |
|
Whoopie
| kampasky: Insount told me. ;) | 13:10 |
|
| kampasky: I'm away for 20 minutes now. CU | 13:11 |
|
| kampasky away too | 13:15 |
| dwmw2_gone → dwmw2_HEL | 13:23 |
| → GeertB joined | 13:33 |
| → meyering joined | 13:52 |
| → luks joined | 13:55 |
| → Newsome joined | 14:10 |
| → Whoopie_ joined | 14:15 |
| → benlau joined | 14:27 |
| → yashi joined | 15:01 |
| → GyrosGeier joined | 15:23 |
| ← yashi left | 15:56 |
| → lyakh joined | 15:57 |
| → alley_cat joined | 17:03 |
| → nud joined | 17:23 |
| → mcr joined | 17:23 |
|
mcr
| hi. if I create a tree using git clone -s, and I later want to pull things into another tree, can I do this without editing the local .git/... alternatives? | 17:24 |
|
| they are "local" on my laptop due to NFS, but the trees in question reference things over NFS, and when I disconnect I want a full tree. Do I perhaps need to do a loopback git+ssh://localhost or something to force things? | 17:25 |
|
| or is there a smarter way? | 17:25 |
| → Tv joined | 17:28 |
| → DrNick joined | 17:47 |
| → krh joined | 17:53 |
| → Eludias joined | 18:08 |
| → arekm joined | 18:41 |
|
arekm
| Can git repository be made available to authenticated users without risk that user will just do rm all files? (normal rm; like over DAV etc loosing all history) | 18:42 |
|
mcr
| maybe. | 18:45 |
|
| do you want them to be able to do updates? | 18:45 |
|
arekm
| yes | 18:45 |
|
mcr
| you can restrict them over SSH to just running git-receive-pack. | 18:46 |
|
arekm
| no other way like via git:// protocol? | 18:46 |
|
mcr
| well, if you don't trust them much, then give each user their own repository, and have a manager responsible for pulling things from them. | 18:47 |
|
| second, realize that git-clone can be very cheap due to hard links, so you can trivially keep as many backups as you want. | 18:47 |
|
| and the backups don't have to be writeable. Also, you could chown all files under .git/objects to a userid other than the user, which would prevent them from over-writing object files. | 18:48 |
|
| (they are essentially only ever written when they are created) | 18:48 |
|
arekm
| I would go to 1) option then - what I'm trying to do is to setup subversion like repository using git (so no one manages it; it just works) | 18:49 |
|
| does git have some kind of rights? like read only to some parts of repository, read-write to other parts etc? | 18:49 |
|
mcr
| not really. you can set up read-only by just not putting users into the group that has write permissions. | 18:50 |
|
| the insides of git do not map to the file system, so you can't provide any restrictions like that. | 18:50 |
| ← kzak left | 18:52 |
|
arekm
| will git work when I'll run git-receive-pack from inetd? | 18:52 |
|
mcr
| no. | 18:53 |
|
| why would you do that? | 18:53 |
|
| if you want to run git-daemon, do so. | 18:53 |
|
| (git-daemon doesn't do writes at all) | 18:54 |
|
| use ssh. seriously. | 18:54 |
|
arekm
| I would have to create chroot, put some kind of pam_sql thing to get ,,virtual'' users. I was hoping that there is more sane way to do that | 18:55 |
| → robin joined | 19:02 |
| → xjjk joined | 19:29 |
| → xl0 joined | 19:37 |
| → xl0_ joined | 19:50 |
| xl0_ → xl0 | 20:22 |
| → robfitz joined | 21:14 |
| → emrys joined | 21:14 |
| → xl0_ joined | 21:37 |
| xl0_ → xl0 | 21:45 |
| → lyakh_ joined | 21:58 |
| → Thumper_ joined | 22:12 |
| → Gitzilla joined | 22:18 |
| → robinr joined | 22:50 |
| dwmw2_HEL → dwmw2_gone | 22:52 |
|
kampasky
| hmmmm | 23:07 |
|
| when arekm comes around next time, please tell him about git-shell and repo.or.cz/w/repo.git | 23:08 |
|
| wrote him mail, nm | 23:43 |
| → treitter joined | 23:57 |
|
treitter
| how do I get a list of the changed/added/deleted files from a given commit? | 23:57 |