IRCloggy #git 2009-10-28

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.

2009-10-28

→ javatexan joined00:02
← juan_arandaalvar left00:02
← tango_ left00:03
jjore Is there a maximum size for a git repo?00:04
← sgh left00:04
jjore I'm kind of worried that my repo might get too big. I'd like to chart space over time but am not sure how best to get a repo's size at a point in time.00:05
cehteh there is certainly *some* limit00:05
but its prolly very high, your patience, money, whatever will run out before that :)00:06
← bdowning left00:06
→ ryoma joined00:06
← tvw left00:06
cehteh really huge git repos becomeing slow00:06
← Ryback_ left00:06
→ bdowning joined00:07
cehteh but look, git can manage the kernel, and some kde folks committed the whole kde project into a single repo00:07
i doubt you do anything biggier00:07
→ schambers joined00:07
cehteh and you should consider split things up into submodules if you plan something that big00:07
jjore So... how big is really huge? I'm up to 900M over approx 1.5 years w/ commits from some 40ish people.00:08
→ jlilly joined00:08
cehteh bigger than some tens of gigabytes00:08
← cubuanic left00:08
jjore But I can't characterize the growth because I don't yet know how to say "the repo on day X was N megabytes"00:09
Ilari Large files destroy the performance much better than huge number of files / huge number of modifications.00:09
cehteh it somewhat depends on your machine too .. some things might get slow or it might need excessive amounts of ram00:09
← javatexan left00:09
cehteh huge binary files which changes frequently are an absolute now00:10
no00:10
few static binary files wont hurt much00:10
← literal left00:10
→ literal joined00:10
cehteh if you want to version your hollyday videos in git you doing something wrong :)00:10
→ lamont` joined00:10
← maio left00:11
← d1b left00:11
← ereslibre left00:11
Ilari Probably first hard limit that would be hit is limit of 4 294 967 295 objects...00:11
jjore I kind of suspect we'd probably have some occasion to store hundreds of megs of data for use by testing code.00:11
I mean, we already do that but we commit that kind of stuff only rarely. The webdevs are probably committing graphics somewhat often.00:12
← gernot left00:12
cehteh i'd suggest to put test-binary data outside .. rsync, tarballs whatever00:12
Ilari I think that storing the data as discrete smaller files as opposed to tarball or so would yield better performance.00:12
→ sgh joined00:12
cehteh (not versioned in git)00:12
wwalker git svn question. I've been doing "git svn clone proto://host/project/branches/some_branch" as the simplest method. Now I am having to switch to other people's branches a lot. there are 70K versions in svn of 40,000 files. I'd like to run a git svn checkout of everything. what command should I run?00:13
or is it even feasible to clone the whole shebang and switch between branches under git-svn?00:13
← ph^ left00:14
wwalker never mind, found it00:14
Ilari The limit at least appears as it being impossible to clone history with >4 294 967 295 objects...00:14
← gasbag left00:15
spearce Ilari: uh, what?00:15
→ ereslibre joined00:15
selckin so thats 40 people doing 3000 commits a day for 10 years ?00:16
cehteh Ilari: on 32bit or generally?00:16
Ilari But at that point, repo size is probably at least something like 1TB if not more... And history having few 100M's of commits.00:16
jjore That's an improbable amount of objects for /humans/ to generate.00:16
So any thought on getting the size of the repo at a point in time?00:17
Ilari The transports only copy one pack, and pack header has 4-byte object count filed...00:17
← jstephenson left00:17
cehteh well 3000 people generating 40 commits a day .. could work :)00:17
→ sgh_ joined00:17
selckin surely you can never have 3k ppl working on the same project-file-set, split it up!00:17
cehteh :)00:17
i dont want to be the poor guy who merges that mess00:18
jjore No, that's improbable too. The communication costs of 3000 people surely prevent them from getting 40 commits a day.00:18
Gah, that'd be horrible.00:18
← sgh left00:18
Ilari And index files alone at that point are something like 110GiB00:18
selckin i've seen some 400 people repos00:18
cehteh and in 10 years the pack headers are 64bit :)00:18
i guess if you approach linus with a reasonable "Hey we hit a git limit by legitimate use here" he will chew out a fix in 1 day :)00:19
← hrehf left00:19
← sgh_ left00:20
← lamont left00:20
← gk left00:22
dbt hold on, if it's 4bn objects that's easier to reaech00:22
imagine a repo of a java project where some util class gets renamed frequently for some stupid reason :)00:23
← tstclair left00:23
← bdowning left00:23
dbt every java file that refers to that class will change.00:23
→ bdowning joined00:23
selckin emperical evidence says nobody has hit it yet, you're probably safe00:24
← KingBug left00:24
→ warthog9 joined00:25
→ EricInBNE joined00:25
→ gk joined00:26
Ilari Hmm... Wonder if deltas >4GiB work correctly (but of course they are insane)00:26
jjore dbt, yeah, but that's a human originated change, to human originated objects, on a human's timescale which tends to not go up to limits like 2**3200:26
→ _ringlej joined00:26
← hobodave left00:26
warthog9 recommends taking the kernel.org log files and each line is a separate commit into a git tree00:27
spearce Ilari: deltas on an object larger than 4 GiB, or an OBJ_OFS_DELTA pointer pointing back over a 4 GiB block of data in the pack?00:27
Ilari spearce: Delta where source is larger than 4GiB.00:27
I have heard some rumours that delta diffing code will crash on those.00:28
→ alezandro joined00:28
← onigiri left00:28
warthog9 Ilari: wouldn't be hard, take a dvd iso and mod it00:28
Ilari Actually, there are easier ways to generate suitable test files (and you can make them take quite small amount of disk space by using holes). The problematic part in testing is large amount (at least 8GiB) RAM it requires).00:29
← justin-g_ left00:29
warthog9 Ilari: true, but at least junio has access to machines that have enough ram for periodic testing00:29
or spearce00:29
← Beirdo left00:30
selckin tries to think of a valid reason to store 4gb files in git00:30
← fhr left00:30
Ilari Still, it shouldn't crash or corrupt data...00:30
selckin true00:30
warthog9 selckin: all of the text of the library of congress in a single file?00:31
selckin i think thats less!00:31
warthog9 selckin: backup copies in the same file?00:31
selckin in 29 languages00:32
← orafu left00:32
→ orafu joined00:32
Ilari Source file: 2^32 0x00s, 65535 0x01s, 1 0x02. Destination: 65535 0x01s, 1 0x02.00:32
Or, actually, destination file can be anything that doesn't contain NULs, and source file is 4GiB of zeroes appeneded with destination file.00:33
The destination file should be 65536 bytes...00:33
← warthog9 left00:33
selckin The print collections of the U.S. Library of Congress: 10 terabytes00:34
oops!00:34
← gk left00:34
→ gk joined00:34
← literal left00:35
selckin A pickup truck filled with books: 1 gigabyte00:35
→ literal joined00:35
Ilari If the delta algo can compress the destination file, something's buggy, because delta offset field has maximum value of 2^32-1.00:35
→ gernot joined00:36
← qhoxie left00:37
← neldoreth left00:37
← vuf left00:37
← elTigre left00:38
← kukks left00:38
← hyperair left00:40
← jlilly left00:40
→ flippo joined00:40
← tfarina_ left00:41
← seanmccann left00:42
→ kpreid_ joined00:42
→ apeiron joined00:42
← ThaDon left00:42
→ ThaDon joined00:43
→ hyperair joined00:43
← agile left00:44
→ Adlai` joined00:45
← apeiron left00:45
← Adlai left00:46
Adlai` → Adlai00:46
→ agile joined00:46
← phantomcircuit left00:47
→ KingBug joined00:48
→ gk__ joined00:48
← EricInBNE left00:49
← markelikalderon left00:50
← gk left00:51
← ThaDon left00:51
← m0 left00:51
→ m0 joined00:52
← yhager left00:53
→ _eric joined00:54
← parasti left00:55
← hyperair left00:56
→ hyperair joined00:57
→ bcardarella joined00:57
→ Swimming_Bird joined00:57
← pointlessjon_ left00:58
→ kardan| joined01:01
← Azeroth-Work left01:01
→ jaysern_ joined01:03
→ jinzhu joined01:04
→ queso_ joined01:04
queso_ How do I move my head back one commit?01:05
→ gretch joined01:05
Adlai queso_: you need to man git-reset01:06
Gitbot queso_: the git-reset manpage can be found at http://git.or.cz/man/git-reset01:06
queso_ Adlai: Thank you :)01:06
← JKac3BEq left01:06
Adlai queso_: if any terminology is unclear, ask away, and maybe google for 'git book'01:06
→ Azeroth-Work joined01:07
queso_ Adlai: thanks :) I should learn a lot more about git because I use it so much, I should drop by here more often.01:08
→ masterloki joined01:09
← _eric left01:12
← spearce left01:12
→ d2dchat joined01:12
← d2dchat left01:13
→ snuxoll_ joined01:14
← jhelwig left01:15
bugfux can I pull in a file from another branch?01:15
sitaram git checkout branch -- file01:16
or git show branch:file > file.new01:16
← kardan_ left01:16
snuxoll_ → snuxoll01:16
← Theravadan left01:18
← schacon left01:19
← relix left01:19
masterloki How can I update an old branch with the master head without mergin my code with the master?01:20
← kennethlove left01:21
→ j416 joined01:21
sitaram masterloki: "merging A with B" is ambiguous. "merging A into B" is clearer, and in this case B changes, A doesn't. Can you rephrase your question?01:21
merging A into B == git checkout B; git merge A # just to be really clear what we mean01:22
masterloki I have and old branch and now master is ahead from me, How can I update my branch withot adding my code to master01:23
sitaram masterloki: git checkout branch; git merge master # this will add master code to branch, not the other way around01:23
Ilari masterloki: 'git checkout <oldbranch>'. 'git merge master'.01:23
→ fualo_ joined01:23
masterloki thanks!01:24
← fualo_ left01:25
j416 masterloki: also read about git rebase, it may be useful01:25
man git-rebase01:25
Gitbot j416: the git-rebase manpage can be found at http://git.or.cz/man/git-rebase01:25
→ hobodave joined01:28
→ jlilly joined01:28
← ryanarn left01:29
→ eno__ joined01:30
← hyperair left01:31
→ dgs__ joined01:33
dgs__ error: Entry 'sqlnet.log' not uptodate. Cannot merge.01:35
fatal: merging of trees c904... and c9600... failed01:35
when trying to merge a branch. As far as i can tell, these files are not important, so happy to get the latest version from the repository, but not sure how I can force that?01:35
hmm. was meant to start that with: "im getting errors like: "01:35
i've tried just deleting the local version, but git pull seems to restore them, and then complain about them again...01:35
← SRabbelier left01:36
← sagsousuke left01:37
→ aliceinwire_ joined01:38
← aliceinwire_ left01:38
← aliceinwire left01:39
→ mrlemao joined01:40
→ dduncan joined01:40
← smuf left01:40
← dduncan left01:40
← eno left01:41
eno__ → eno01:42
← mrlemao left01:43
← flazz left01:44
→ hyperair joined01:45
→ jhelwig joined01:47
← bugfux left01:48
→ WALoeIII_ joined01:49
→ mdemmitt joined01:50
← WALoeIII_ left01:50
→ WALoeIII_ joined01:51
← mjf left01:52
→ mncaudill joined01:54
← icwiener left01:55
← WALoeIII left01:58
WALoeIII_ → WALoeIII01:58
← Yuuhi left01:59
← ereslibre left02:00
→ aliceinwire joined02:01
← Azeroth-Work left02:02
→ Azeroth-Work joined02:02
→ Yewbacca2 joined02:02
← Yewbacca left02:02
← WALoeIII left02:02
← ruphy left02:03
→ ruphy joined02:03
→ Yewbacca joined02:04
→ VVSiz_ joined02:04
← MisterN left02:04
→ MisterN joined02:04
← Yewbacca2 left02:05
→ abms1116 joined02:06
→ shadowmaster joined02:06
← abms1116 left02:06
→ Adlai` joined02:07
→ abms1116 joined02:07
← luke-jr left02:07
→ luke-jr joined02:08
← jlilly left02:11
← Adlai left02:11
Adlai` → Adlai02:11
← Chillance left02:12
← snuxoll left02:12
→ cao joined02:19
cao → Guest6204002:20
→ snuxoll joined02:20
→ JKac3BEq joined02:21
← VVSiz left02:22
→ krynnotaur joined02:23
← krynnotaur left02:24
← __iron left02:25
← kaseano left02:25
→ __iron joined02:26
← Guest2583 left02:26
← godsflaw left02:29
→ WALoeIII joined02:30
→ javatexan joined02:31
→ upul joined02:31
← javatexan left02:31
queso_ Today a co-worker made a commit that broke things. I reset my head to just before his commit. But now I've made modifications and want to make a commit. What can I do to prevent my commit from attempting to merge with his, so that it ignores his commit and basically ignores it?02:31
← flippo left02:32
queso_ (ignores his commit and basically ignores it = it's been a long day ;)02:32
← eno left02:32
→ juan_arandaalvar joined02:33
Ilari queso_: Either force the push (dangerous) or merge his commit using '-s ours'.02:33
queso_ Ilari: what does '-s ours' mean?02:33
Ilari queso_: Effectively mark commits as merged, but don't merge any changes in.02:34
queso_ Ilari: okay, thanks02:34
→ loops joined02:35
← jepler left02:35
→ jepler joined02:36
→ eno joined02:37
→ archis joined02:37
← archis left02:38
← shadowmaster left02:39
← juan_arandaalvar left02:43
→ Adlai` joined02:43
← Adlai left02:44
Adlai` → Adlai02:44
← mncaudill left02:44
→ Hideki_ joined02:46
→ Theravadan joined02:47
→ mncaudill joined02:47
→ fynn joined02:49
fynn Good evening.02:50
Is there a way to reset/checkout just the deleted files in a directory, but not the modified files?02:50
revans → revans|away02:50
→ flippo joined02:51
loops fynn, something like git checkout -- $(git ls-files -d)02:52
→ krynnotaur joined02:52
→ _VVSiz_ joined02:52
fynn loops: thanks02:53
← DrTrevorkian left02:54
fynn yeah that works verbatim.02:56
Ilari Seems bit dangerous if filenames have spaces...02:56
← jschoolcraft left02:57
loops might be nice if ls-files had shell quoting option02:58
Mikachu just put it in quotes02:59
oops, i should read before talking :)02:59
revans|away → revans03:00
← Hideki_ left03:01
→ gko joined03:03
→ gko` joined03:03
← revans left03:03
← bcardarella left03:03
← jinzhu left03:07
← hyperair left03:08
← VVSiz_ left03:10
→ rchavik joined03:11
← Rovanion left03:11
← gko` left03:17
← gko left03:17
→ cytrinox_ joined03:20
→ secoif joined03:26
← WebDragon left03:26
← aziz left03:28
← jaysern_ left03:28
→ WebDragon joined03:28
← squentin left03:28
→ jeffmoss joined03:29
tk sitaram: dont suppose you're still around?03:30
← richcollins left03:31
→ hgimenez joined03:32
sitaram tk: still? my day has just started!03:34
it's 9:04 am and I'm at work (India!)03:34
→ argonel joined03:34
tk sitaram: ohh, which part? my work has a team over there actually :P03:34
sitaram so go ahead; as I said when I'm at work there's always the chance of lag but otherwise I'm here03:34
Hyderabad :)03:34
lucs YEah, sometimes he works :)03:34
tk sitaram: ohh I was just looking for better info on post-commit hooks, want to push files to a dev server after commits for in progress projects03:35
so far google hasnt turned up anything useful but still digging03:35
← cytrinox left03:36
cytrinox_ → cytrinox03:36
sitaram tk: post commit hooks... why not post-push (aka post-update) so these hooks are only in the server, not on dev workstations? Or do you *need* them in dev workstations?03:37
tk sitaram: post-commit --- svn :P you are right, update is the appropriate one for git03:37
hmm there isnt even a sample for that one03:38
← ia left03:38
tk sitaram: our guys are in New Delhi03:39
→ hyperair joined03:39
← mdemmitt left03:39
sitaram tk: post-update is easy, wait03:40
← fynn left03:41
→ dunolie_ joined03:42
← schmukkk left03:43
sitaram tk: http://toroid.org/ams/git-website-howto might be a good start (although now I notice he's using post-receive; because that apparently supercedes post-update now)03:43
→ schmukkk joined03:44
← dunolie left03:44
dunolie_ → dunolie03:44
← hgimenez_ left03:44
→ Dashkal joined03:46
→ cedricv joined03:47
→ ia joined03:48
← aliceinwire left03:51
← hobodave left03:51
← legnaleurc left03:52
← hyperair left03:53
→ hyperair joined03:54
→ bcardarella joined03:57
→ jrmuizel joined03:58
tk sitaram: thanks, will take a look03:59
← scarabx left04:02
← flippo left04:07
← schmukkk left04:07
← tjafk left04:09
→ tjafk joined04:09
→ ericflo joined04:11
← modsognir left04:11
ericflo How do I make a commit that reverts all the changes I've made since revision X?04:12
I just want to rewind.04:12
→ modsognir joined04:12
→ richcollins joined04:12
← hyperair left04:14
← jfkw left04:15
→ urbanmonk joined04:16
← bdimcheff left04:17
cehteh ericflo: git revert04:18
ericflo cehteh: Seems that reverts a specific revision.04:18
jrdavid how about: git reset --hard version04:19
cehteh uhm04:19
ericflo: is that repo already published or private?04:19
ericflo git status gives me "Your branch is behind 'origin/master' by 11 commits, and can be fast-forwarded."04:19
cehteh: nope, it's published04:19
cehteh then not reset --hard04:19
you can revert with -n and stack reverts together04:20
ericflo yeah, I just git reset --hard'ed to get back to HEAD again04:20
→ schmukkk joined04:20
← schmukkk left04:20
ericflo cehteh: That looks like it will work, but I have to do that for every medium inbetween HEAD and the one I want?04:21
← urbanmonk left04:21
← dunolie left04:22
cehteh i just think there is prolly some easier way04:22
reset or checkout but not hard and make that a new commit04:22
→ urbanmonk joined04:22
cehteh i never had to undo multiple commits yet :)04:22
engla git diff HEAD..HEAD~10| git apply04:22
just apply a reverse diff04:23
cehteh git checkout HEAD~10 -- . maybe but not sure04:23
ericflo engla: Yeah! That worked. Thanks a bunch.04:23
cehteh 1000 ways to do something :)04:24
← gretch left04:24
→ gretch joined04:24
ericflo :D04:24
thanks for the help though, cehteh04:24
jrdavid thinks engla is clever04:24
← bcardarella left04:25
→ hyperair joined04:25
engla has had a lot of git today so I'm lucid04:25
→ bcardarella joined04:25
jrdavid What's a good way to always --decorate git log?04:27
← ericflo left04:27
engla an alias?04:28
→ gretch_ joined04:28
← gretch left04:28
→ DavidKlein joined04:29
engla if you create an alias that you'll always use for viewing the log, you can throw in some other options as well, like --stat04:30
← DavidKlein left04:30
→ schmukkk joined04:30
jrdavid yeah I did that, but I always end up typing 'git log' instead of the alias (habit)04:32
engla I only have alias habits, don't know if that is better. well, you can't override log and I don't see any config option04:34
override with an alias I mean04:34
→ SRabbelier joined04:36
← urbanmonk left04:36
jrdavid yeah I looked for log.default_options too...04:37
← TML left04:38
→ TML joined04:38
→ dunolie joined04:39
→ Adlai` joined04:40
← Adlai left04:41
Adlai` → Adlai04:41
→ urbanmonk joined04:41
→ jlilly joined04:42
← patcito left04:42
← urbanmonk left04:43
→ urbanmonk joined04:44
→ schacon joined04:48
← cworth left04:51
← ryoma left04:54
→ javatexan joined04:56
→ zoonfafer joined05:01
← JoeConyers left05:06
← javatexan left05:07
→ JoeConyers joined05:07
← solarbunny left05:10
→ gaveen joined05:10
→ bsdvodsky joined05:11
→ murph joined05:13
← jspiros left05:15
→ yhager joined05:16
→ scientes joined05:18
scientes how do i remove new files?05:18
git reset --hard doesnt do it05:18
→ queso__ joined05:19
← murph_ left05:19
← Bass10 left05:20
→ MariusAZ joined05:20
MariusAZ How do you apply just a patch file to git? is it something like cat patch.file | git apply?05:21
← madewokherd left05:21
crab or git apply patch.file05:21
MariusAZ oh ok05:21
That's what I thought, thanks05:21
← MariusAZ left05:21
→ jaysern_ joined05:22
→ JohnFlux joined05:22
← masterloki left05:23
→ jpic_ joined05:23
JohnFlux I still haven't worked out how to git pull exactly..05:23
if I do: git pull origin branchname what happens?05:23
it will pull the local branch called "branchname" ?05:23
← jaysern_ left05:23
JohnFlux or the remote branch called "branchname" ?05:23
jpic_ hi, i think i confused git again, i'm trying to add some files but it's "not doing anything": files are not listed in git status05:23
→ jspiros joined05:23
→ mdemmitt joined05:24
→ jaysern_ joined05:24
← jaysern_ left05:24
engla JohnFlux: remote. perhaps it's easier if you first download all remote changes "git fetch origin" and you can inspect them in peace before you merge or rebase to integrate the changes05:24
→ Tuomas joined05:25
JohnFlux engla: okay so it pulls the remote branch called "branchname" to where? the current branch that I have checked out?05:26
jpic_ maybe because that directory i' mtrying to add has a .git subdir of its own?05:26
→ a389742 joined05:28
engla yes, exactly05:29
that's the danger of pull05:29
jpic_: might be that, I know 0 about submodules05:29
JohnFlux engla: hmm. and if I do: git push origin branchname. What does that do?05:30
← jrmuizel left05:31
JohnFlux engla: push the _current_ checkout to remote branch "branchname" ? or push the local checkout "branchname" to remote branch "branchname"?05:31
← queso_ left05:32
engla the latter05:34
push and pull are not symmetric05:34
→ queso_ joined05:34
← queso__ left05:36
→ jrmuizel joined05:36
← schacon left05:39
← loops left05:39
← SRabbelier left05:40
→ FunkeeMonk joined05:43
← __iron left05:44
→ abe_t joined05:45
jpic_ apparently i should add a submodule entry in my superproject with a path and url05:45
→ roop joined05:45
→ jceb joined05:46
← abe_t left05:46
→ sprasanna joined05:46
jpic_ it seems like this has to be done *in every single clone*, unlike svn:externals05:48
hyperair hmm? no, it isn't required for every clone05:48
jpic_ my bad :)05:48
← Swimming_Bird left05:49
← merlin83 left05:53
→ maleknet joined05:56
→ macmaN6789 joined05:57
← Theravadan left06:01
→ fualo_ joined06:01
← mdemmitt left06:03
→ murph_ joined06:04
JohnFlux engla: seems so confusing - you'd think push and pull would be symmetric06:05
xenoterracide_ → xenoterracide06:05
← jrmuizel left06:06
← vinse_ left06:07
← abms1116 left06:09
← murph left06:09
← d3z left06:09
← bcardarella left06:09
← hyperair left06:10
→ jrmuizel joined06:10
← jrmuizel left06:10
→ hyperair joined06:10
← nighthwk1 left06:12
→ nighthwk1 joined06:12
← nighthwk1 left06:12
← karihre left06:13
→ karihre joined06:14
→ _jason553839 joined06:17
→ Axius joined06:19
← johan-s left06:21
← jceb left06:26
_VVSiz_ → VVSiz06:27
← Axius left06:28
→ Axius joined06:28
← agenteo left06:33
← lamont` left06:35
Mikachu JohnFlux: push and fetch are closer06:35
← engla left06:36
→ ph^ joined06:36
← _jason553839 left06:39
→ roop_ joined06:41
← JackWinter left06:41
← niki left06:42
Phurl is there any work on a libgit?06:42
i would like to call commit from a webserver, but do not want to use exec because of security.06:43
← Aestas left06:43
← jeffmoss left06:44
→ kaseano joined06:46
→ queso__ joined06:50
→ priidu joined06:50
Mikachu that doesn't really make much sense, since commit execs a lot of stuff06:52
← eletuchy left06:52
← hyperair left06:54
→ hyperair joined06:54
→ psoo joined06:56
← psoo left06:56
← roop left06:57
→ jinzhu joined06:59
→ queso___ joined07:00
→ cworth joined07:01
← FunkeeMonk left07:01
← queso_ left07:03
→ abms1116 joined07:03
→ clayjar joined07:03
← maleknet left07:03
→ alireza_ joined07:03
clayjar G'evening folks.07:03
Wow, a huge crowd.. I guess this is normal here.07:04
vkroll good morning clayjar07:04
→ apm121 joined07:04
clayjar Anyway, here is my question: I've installed and set up git, gitosis, and gitweb (through Apache). I clone gitosis-admin from my git server (repo), and then update the gitosis.conf and then commit -> push origin master; shouldn't the changes be immediately effective?07:05
It seems like I have to run sudo -H -u git gitosis-init < id_dsa.pub every time I make a change to make it effective.07:05
apm121 Is something like git-rus irc or maillist exists? Having problem with english, want russian ....07:06
clayjar I'm not sure if there is a Russian language FreeNode, but my understanding is that most of FreeNode users are English speakers...07:06
← reidab left07:07
clayjar Is everyone really asleep or busy typing answers??07:07
crab you certainly shouldn't need to run gitosis-init each time07:08
sitaram clayjar: that might mean your post-update hook isn't chmod 755 or something07:08
clayjar crab, true.07:09
sitaram clayjar: http://sitaramc.github.com/the-list-and-irc/gitosis-troubleshooting.html may help; I have to run for a meeting now sorry07:09
clayjar sitaram, tx for the reference, c ya.07:09
→ roop__ joined07:09
clayjar well, post-update is 755 all right.. I've checked it every time just to make sure.07:09
→ sde_antrix1 joined07:09
→ Weasel[DK] joined07:10
← roop_ left07:10
clayjar by the way, is post-update normally symlinked to some other location?? I think when I first encountered it it was symlinking to an non-existent file, so I deleted it and copied post-update.sample to post-update under hooks/ folder.07:10
← apm121 left07:11
→ nothingmuch joined07:12
← nothingmuch left07:12
→ ludde joined07:12
→ marh joined07:13
← queso__ left07:13
→ nothingmuch joined07:13
→ Grahack joined07:14
marh hi! i'm trying to put a debian installation (on my openmoko phone) under git. adding stuff worked but now there are ~200k loose objects and git cg tells me 'nothing new to pack'. can someone give me a hint on what's wrong? or shouldn't i use git for this in the first place?07:15
← voker57 left07:16
marh (it is < 512M in size)07:16
← gretch_ left07:17
→ nitrate joined07:18
← nitrate left07:18
→ saurabhverma joined07:18
→ thiago_home joined07:18
j416 marh: are the loose objects unreferenced?07:25
in that case git prune would kill them.07:26
→ ixti joined07:27
→ gretch_ joined07:27
← secoif left07:29
marh j416: ah.. i wasn't exactly sure what you mean by that.. but i committed and it seems to work now :)07:29
← Axius left07:31
j416 marh: 'git prune' deletes objects that are not referenced = unreachable07:32
it has nothing to do with your committing or not07:32
it is just for cleaning07:32
→ drizzd joined07:33
clayjar hook/post-update doesn't seem to be running properly... thus, no immediately update for gitosis07:33
← JoeConyers left07:35
→ khmarbaise joined07:35
→ babilen joined07:36
→ fualo__ joined07:37
clayjar anyone familiar with this issue?07:37
→ bentob0x joined07:37
← Rolenun left07:38
← segher left07:39
sitaram ls -al the file and see if it has +x07:40
oh wait I just saw what you wrote up there07:40
← bambule left07:40
→ arekm joined07:41
← khmarbaise left07:42
→ elTigre joined07:42
clayjar I su 'd into git, and then tried to run exec git-server-update-info.. it returns: exec: 11: git-update-server-info: not found07:46
the only way gitosis would update itself is when I manually run "sudo -H -u git gitosis-init < id_dsa.pub"... I wonder why the hooks wouldn't be operating properly...07:46
← fualo_ left07:47
← selckin left07:47
→ galderz joined07:47
arekm is looking for some software like gforge, fusionforge that supports svn and git - does anyone know such beast? (smaller->better)07:49
← roop__ left07:49
→ roop__ joined07:50
→ scientes_ joined07:51
← davido left07:51
← Nugget_ left07:51
→ Nugget joined07:51
→ schmukkk2_ joined07:51
→ davido joined07:51
← schmukkk left07:52
← Dashkal left07:52
→ mithro joined07:54
→ dpino joined07:56
← Adlai left07:57
← hyperair left08:00
← arekm left08:00
→ khmarbaise joined08:00
→ cjs joined08:02
→ ntoll joined08:03
← khmarbaise left08:05
→ hyperair joined08:05
→ zoniq joined08:06
→ psankar joined08:06
cjs So if you had a repo you declared as the "master," to which everybody pushes, and from which you roll out your production software, you could do a hot backup of this pretty easy by just having a post-update hook that pushed the changes to the hot backup server, right?08:06
← scientes left08:07
← jlilly left08:12
thiago_home yes08:12
→ jschoolcraft joined08:14
← gretch_ left08:15
← jschoolcraft left08:15
→ jschoolcraft joined08:16
→ tschundeee joined08:17
← RobertLaptop left08:20
← mithro left08:22
→ agkawj joined08:22
← Grahack left08:23
→ merlin83 joined08:25
→ cilly joined08:25
→ gretch_ joined08:26
← scientes_ left08:26
→ scientes_ joined08:26
← schmukkk2_ left08:27
→ chamo joined08:28
chamo hi all08:28
→ schmukkk2_ joined08:28
→ voker57 joined08:29
→ DerKomische joined08:29
→ neldoreth joined08:30
neldoreth → Guest6081708:30
→ fluxusps joined08:31
← chamo left08:31
→ chamo joined08:31
← bsdvodsky left08:31
Guest60817 → neldoreth08:32
← thiago_home left08:32
→ niki joined08:33
multani0 → multani08:33
→ shruggar joined08:33
→ tobi joined08:34
→ SeySayux joined08:35
SeySayux Hi, how do I clone all branches from a remote?08:35
→ jfkw joined08:35
→ gusl joined08:36
→ queso_ joined08:36
gusl hi08:37
I'm working with github as a way to back up my projects08:38
but I seem to have forgotten the procedure for committing and pushing things there08:38
shruggar SeySayux, what is your goal? iirc, the default is to copy all refs from the remote, though local branches will only be created for the default08:38
Rhonda SeySayux: When you clone you get all branches already. "git branch -a" lists them all.08:39
→ tobi_ joined08:39
gusl I thought the procedure was: (1) edit file.txt (2) git add file.txt (3) git commit -m 'backup' (4) git push origin master08:40
is this missing a step?08:40
← hyperair left08:41
→ hyperair joined08:41
shruggar gusl, that looks right for the general case, I don't know about github, however08:42
→ tango_ joined08:42
gusl shruggar: so I don't need to remote add again?08:42
← tobi left08:43
SeySayux Hmm, I only get 'master' and a lot of remote/origin stuff -- how do I get that in my local repo?08:43
shruggar gusl, "remote add" just adds configuration information about the remote, giving it the name you specify. Once it's added, it gets referred to by that name from then on (in this case, "origin"). So, no, you don't need to add a remote every time (unless you haven't already)08:43
SeySayux, remote/origin _is_ in your local repos, it's just a reference to something which came from remote/origin08:44
gusl thanks, shruggar08:44
Rhonda SeySayux: git checkout -b somebranch remote/origin/somebranch - read man git-checkout08:44
Gitbot SeySayux: the git-checkout manpage can be found at http://git.or.cz/man/git-checkout08:44
chamo sitaram: are you around?08:45
gusl and indeed, this seems to work! It only doesn't work when I use the shell inside emacs. :-(08:45
shruggar damn hippies! :)08:45
→ aatanasov joined08:46
sitaram chamo: yes08:46
gusl in my emacs shell, I can do git commit, no problem. But when I try to do git push, it doesn't do anything... no asking for my passphrase.08:46
yhager → yhager|away08:46
→ ArdRigh joined08:47
gusl I'm wondering if the emacs shell knows the ssh key08:47
shruggar gusl, that's probably got something to do with the way things redirect (though I don't know about emacs specifically). Try setting up ssh-agent, if github supports it.08:47
chamo sitaram: can you help me to configure gitolite? I got an error when trying to get my repo woth fatal exceptions08:47
gusl ssh-agent on the emacs shell gave me some stuff like SSH_AUTH_SOCK=... SSH_AGENT_PID=...08:48
← hyperair left08:48
← DerKomische left08:48
→ hyperair joined08:49
→ thaz joined08:49
← queso___ left08:49
→ ph^_ joined08:49
sitaram chamo: sure --- I will be back in 5 minutes... hope thats ok (am at work!)08:50
chamo sitaram: okie dokie im waiting08:51
trying to figure out08:51
→ eletuchy joined08:52
← gusl left08:52
→ funkmonk joined08:54
← Pupeno left08:55
→ johan-s joined08:57
← tobi_ left08:58
→ Pupeno joined08:58
→ RobertLaptop joined08:58
← eletuchy left08:59
→ eletuchy joined08:59
→ jstephenson joined09:00
← babilen left09:01
clayjar fixed that hooks problem... I had git 1.6.5.1 installed with the older version of gitosis from Ubuntu Jaunty repository.. they weren't compatible...09:04
I uninstalled gitosis, and then git clone from Tom's repo, then did a reinstall. Now, it works like a charm.09:04
→ struberg joined09:06
clayjar gitosis should have more documentation, imho.09:07
SeySayux Hi, how do I turn all tracking branches into local branches?09:07
sitaram clayjar: gitosis is obsolete09:07
err.. stable09:07
thats the word :)09:07
clayjar http://sitaramc.github.com/0-installing/9-gitolite-basics.html and http://github.com/sitaramc/gitolite09:07
clayjar gitolite, eh? Hmmm, does it have a seamless integration with Git?09:08
← ArdRigh left09:08
sitaram clayjar: I have no idea what that means09:08
clayjar I'll just read the page and see if it is or not...09:09
sitaram or you could help me understand what you mean by seamless integration :)09:09
clayjar it basically means working as "one body/unit."09:10
← hyperair left09:10
sitaram clayjar: thank you; but I did not mean that :)09:10
clayjar: actually forget it; if you were ok with gitosis, gitolite will be fine too, regardless of what you mean by "seamless integration" :)09:11
→ mattayers joined09:13
→ psoo joined09:13
→ hyperair joined09:14
→ tono joined09:14
← krawek_ left09:15
clayjar sitaram, I'm not sure what you are trying to achieve with gitolite... gitosis simply utilizes what you are labeling "ssh magic" (technically, it's called OpenSSH public key authentication) to only allow git push/pull.. you don't need to create unix accounts for multiple users. But you seem to have done a good job of creating an alternative way to do the same thing... I hope it's as secure...09:16
...as Gitosis...09:16
sitaram clayjar: read the bottom of http://github.com/sitaramc/gitolite/blob/pu/README.mkd09:16
clayjar Thanks, this is much clearer... it's impressive how you've managed to rewrite it to allow for branch-level access.09:17
sitaram clayjar: also read http://sitaramc.github.com/0-installing/9-gitolite-basics.html -- which is much more detailed description of what you just said; it has been useful to help people overcome ssh problems09:17
roop__ → roop09:17
sitaram clayjar: nothing is impressive once it is documented, and I document all of it :)09:17
clayjar Sitaram, nice. It'd be even better if you could find someone to help you clean up the document a bit, but it looks like you've produced a very useful utility out of gitosis.09:19
→ ph^__ joined09:19
sitaram thanks... well if I weren't at work right now I'd ask what needs cleaning up but I;m swamped right now :( so maybe later...09:20
→ ph^___ joined09:20
clayjar It's 3:21 am here.. I think I'm little overworked.. needs to get some shuteye for at least for few hours... talk to you later.09:21
← abms1116 left09:22
SeySayux Hi, how do I turn all tracking branches into local branches?09:23
sitaram one by one09:23
clayjar git fetch origin ?09:24
sitaram (unless it's a bare repo, in which case you *could* use --mirror but that has other implications I'm not sure of)09:24
Mikachu git fetch . 'refs/remotes/origin/*:refs/heads/*' should do it, i think09:24
Mikachu leaves others to explain how to recover lost commits with the reflog09:25
sitaram Mikachu++ -- didnt think of that :)09:25
doener detach HEAD first09:25
and that won't lose commits, there's no force involved09:25
SeySayux Is there a way to already do this with git-clone? I can't find anything in the manual about it...09:26
Mikachu SeySayux: why do you want to do this?09:26
clayjar SeySayux, could deleting the directory and recloning help? If you're just talking about getting replacements.. sorry if it sounds too simple, but I may be misunderstanding your question.09:26
SeySayux Erm, because I want to have all branches?09:27
Mikachu but what are you going to do with them?09:27
SeySayux modify them?09:27
Mikachu you may proceed then09:27
← sitaram left09:28
→ sitaram joined09:28
SeySayux Mikachu: hmm, doesn't seem to work...09:29
doener SeySayux: there's no such option to clone, because usually, you create branch heads "on demand", and have only those that you currently care about09:29
SeySayux hmm, I care about them all... but I found some stuff in the GUI that can help me... thanks anywayus09:29
s/us$/s/09:30
doener SeySayux: you work on all branches at the same time?09:30
impulze who doesn't :P09:30
← ph^ left09:32
← xenoterracide left09:32
← fualo__ left09:33
← hyperair left09:33
← priidu left09:33
clayjar I'm a greenhorn w/ git (only a week now), but I'm still unclear (probably inexperienced) with how all the branching is supposed to work with centralized repo...09:33
← ph^_ left09:34
→ Sho_ joined09:34
→ xenoterracide joined09:34
clayjar I guess I need to eat my fill of mistakes...09:34
← Pupeno left09:34
→ trochala joined09:35
← xenoterracide left09:35
← ph^__ left09:36
→ xenoterracide joined09:37
← xenoterracide left09:38
← psankar left09:38
→ psankar joined09:39
→ __iron joined09:39
rchavik is there a way to make git 'prune' a repository? ie, simply forget or delete the history older than one year ago09:41
→ xenoterracide joined09:41
doener rchavik: not without rewriting the new history as well09:42
← clayjar left09:42
← xenoterracide left09:42
rchavik doener: ok.. thanks, i'm a newbie.. so i'm going to take that as a no :)09:43
doener rchavik: different question: Why would you want to do that?09:44
rchavik anyway.. it's going to be some time before my repo get big ;)09:44
doener: well at some point, i simply won't care for old revisions...09:45
→ aliceinwire joined09:45
rchavik eg: the repo is for website.. and old theme files can simply be deleted09:46
→ xenoterracide joined09:46
doener rchavik: I regularly find myself reviewing very old commits to see how some code came into existence, so I can understand how a bug (that didn't even exist back then) was caused and how to fix it09:46
→ Grahack joined09:47
rchavik doener: well.. in my old case, old themes may contain obsolete images..09:49
doener: mostly concerned with old/unused images in themes that are not being used anymore in git repo..09:50
→ KyleMac joined09:52
→ JackWinter joined09:52
KyleMac what's the quickest way to find and recover a file that I've somehow lost but I know is in the respository somewhere within at least the last 5 commits?09:53
doener KyleMac: do you know the file name?09:53
KyleMac yeah09:53
doener KyleMac: git log -1 -- <name> # Finds (one of) the last commit(s) in your checked out branch that had the file09:54
KyleMac: then "git checkout <commit> -- <name>" to get the file from that commit and put it into the index and working tree09:54
KyleMac: (well, it actually finds one of the last commits that _modified_ the file, but if it was deleted, it's already the last one that had it ;-))09:55
KyleMac I can find it but when I try to check it out it says the filename did not match any files known to git09:56
doener KyleMac: ah, d'oh. The found commit deleted the file, so it doesn't have it09:57
KyleMac: use "git checkout <commit>^ -- <name>" instead, to the get file from the first parent of the commit that deleted the file09:57
→ gunmuc joined09:58
KyleMac that didn't work either, but git log -2 gave me the commit to use09:58
thanks09:59
→ Juice10 joined10:00
← Juice10 left10:00
KyleMac btw does anyone have a solution for a weird problem with git I have on windows. When I try to push somewhere, I get the "no authorisation method" or whatever thing, which I can solve by setting GIT_SSH. But if I reboot the error comes back and I solve it by deleting GIT_SSH, but then if I reboot it comes back and I solve it...10:02
it doesn't matter whether I use the thing that came with msysgit or putty10:03
→ Redworm joined10:03
→ relix joined10:03
← Redworm left10:03
→ abms1116 joined10:03
→ tvw joined10:04
→ spaceonline joined10:07
← merlin83 left10:09
← kaseano left10:09
→ fualo_ joined10:10
→ capndiesel joined10:15
→ jayvee joined10:16
jayvee Quick Q that I couldn't find a straight answer to on the intertubes:10:17
How do I do a lightweight checkout of a git branch?10:17
i.e. download just the latest code without downloading 100MB of worthless buggy crap from 3 years ago10:17
In bzr and svn it is very easy: just use the ‘export’ or ‘checkout’ command.10:18
← j416 left10:19
doener jayvee: "git archive", but often repos don't allow that to be used remotely.10:19
DrNick git clone --depth10:20
jayvee doener: it's from GNOME. hopefully they will.10:20
doener jayvee: there's also "git clone -depth N", but read the man pages comments on that10:20
→ siks joined10:20
jayvee DrNick: that is cool, I wish bzr had that ability.10:20
so --depth=1 might fulfil my needs?10:21
doener: I can't use git archive with a remote repository.10:22
doener jayvee: maybe, maybe not. Depth-limited clones have some restrictions10:22
jayvee My whole goal is to save bandwidth. (I'm a poor deprived Australian.)10:22
← balor left10:22
doener jayvee: and I'd not expect the space saving to be very big10:23
jayvee *cries*10:23
Well I'm doing it with --depth=1, and it looks like it'll only need to download about 10MB.10:24
oh there we go10:24
did it in only 2.8310:24
cheers doener and DrNick10:24
doener jayvee: for empathy: full clone: 9.37MB downloaded, --depth 1 clone: 5.82MB.10:26
jayvee heh, that's not much different for empathy10:27
← roop left10:27
doener jayvee: given how much less useful the depth=1 clone is (unless you're basically just interested in a snapshot, and don't look at the code at all), I'd rather get those extra 4MB10:27
jayvee yeah k10:28
but there is a repo I'm planning to work with that has lots of crufty history10:28
→ intripoon joined10:28
jayvee binary files being added and removed over time10:28
doener ouch10:29
← jxavier_ left10:30
→ khmarbaise joined10:30
jayvee like10:30
hundred meg SQL dumps10:30
doener sounds like someone forgot to cleanup the history while doing the import into git10:31
(I doubt that those SQL dumps are actually used, are they?)10:31
jayvee well, they were at one point10:32
but are horribly outdated10:32
and then "deleted"10:32
by someone that clearly doesn't understand you shouldn't put that stuff in a VCS10:32
→ fualo__ joined10:34
← tango_ left10:36
← eletuchy left10:38
→ eletuchy_ joined10:39
→ kkoehne_ joined10:39
← Creap left10:39
→ Creap_ joined10:40
← kkoehne left10:40
→ Axius joined10:40
→ lamont` joined10:40
→ merlin83 joined10:40
→ ryah joined10:43
ryah how do i do this more easily: "git diff b893859c34f05db5c45f416949ebc0eee665cca6^..b893859c34f05db5c45f416949ebc0eee665cca6"10:43
that is, see what changed in one commit10:43
Pieter git show b893859c34f05db5c45f416949ebc0eee665cca610:43
ryah great thanks10:43
← fualo_ left10:44
→ hyperair joined10:45
← jayvee left10:45
shruggar or git log -p b893859c34f05db5c45f416949ebc0eee665cca610:47
and of course, git show b8938510:47
funkmonk i want to put a config file into my project. there should be a dummy of it in every version, but i want my version with my secrets staying in my working folder. is that even possible? my config file version would always make the repro dirty, right?10:48
← dsch04 left10:48
vkroll funkmonk: you can put your private file in .gitignore10:48
funkmonk yeah but that doesnt work after committing it.10:49
vkroll no, but can't you require your private file in the offical one and put the private in your .gitignore?10:50
funkmonk and I need to commit it, so the project is complete and others get the config file with dummy input10:50
how would people downloading the project from github get the file then?10:51
vkroll in perl you can do Things like if( -e my private_file) { ## use it } else { use the offical one }10:52
funkmonk hm. thing is it's an xcode project. but i will investigate that10:53
← VVSiz left10:55
→ priidu joined10:56
→ aziz joined10:56
yhager|away → yhager10:58
← queso_ left10:58
← khmarbaise left10:58
← richcollins left10:58
→ VVSiz joined11:02
→ queso_ joined11:03
→ khmarbaise joined11:04
← alireza_ left11:05
→ b0fh_ua joined11:05
Creap_ → Creap11:05
charon funkmonk: i usually track the file anyway, and have separate config branches with the "real configs". then a 'git checkout master^0; git merge realconfig' makes a temporary merge that is testable. remember to 'git checkout master' afterwards to reattach your HEAD11:06
b0fh_ua Hi there! We would like to migrate from SVN to git, however we still have to commit to remote SVN repository. There are several accounts used by developer for committing to SVN, the question is - how can we make it work in the way when all commits are going into git repository, then after review these changes are pushed to SVN, but each commit to svn is being made by appropriate developer (who actually did that commit)?11:06
→ alireza_ joined11:06
→ aliceinwire_ joined11:08
funkmonk charon: ah clever. thanks!11:08
← Axius left11:09
← SeySayux left11:11
← aliceinwire_ left11:12
← eletuchy_ left11:12
→ eletuchy joined11:12
→ alireza__ joined11:15
← upul left11:15
→ engla joined11:15
→ hingwah joined11:17
→ brizly1 joined11:17
→ tango_ joined11:17
← engla left11:18
← priidu left11:18
← tschundeee left11:20
→ fualo_ joined11:21
→ tek0_ joined11:22
← alireza_ left11:23
← tek0 left11:23
→ Axius joined11:23
← xenoterracide left11:23
→ ArdRigh joined11:24
→ lrojas joined11:25
→ xenoterracide joined11:25
← eletuchy left11:26
ryah → ryah_away11:27
← xenoterracide left11:28
← fualo__ left11:28
← lrojas left11:29
→ xenoterracide joined11:29
← fr left11:29
→ fr joined11:29
→ Yuuhi joined11:31
→ tek0 joined11:31
← tek0_ left11:33
← brizly left11:33
← _ringlej left11:34
← jepler left11:37
← psankar left11:38
→ corni joined11:38
→ temoto joined11:38
→ trochala` joined11:39
→ jepler joined11:41
← temoto left11:42
← agkawj left11:43
← khmarbaise left11:44
→ Ryback_ joined11:49
→ _jason553839 joined11:49
→ holizz_ joined11:50
← fujin left11:52
→ fujin joined11:52
← fualo_ left11:52
← trochala left11:53
← funkmonk left11:54
→ catphish joined11:54
→ boto joined11:54
catphish → adam11:55
adam → catphish11:55
← Axius left11:55
← catphish left11:55
← tango_ left11:56
b0fh_ua 15+133+124.611:56
→ engla joined11:56
← b0fh_ua left11:56
← trochala` left11:57
Creap so I moved a repository with rsync and it kind of works, but I am using git-svn and I get "Unable to determine upstream SVN information from working tree history"11:57
should I run some command to "init" git-svn again?11:57
→ lorandi joined11:58
← doener left11:59
→ doener joined11:59
→ maattd joined12:01
→ d0k joined12:03
→ queso__ joined12:04
→ fualo_ joined12:04
← queso_ left12:06
← ArdRigh left12:06
← abms1116 left12:07
→ froschi joined12:08
← a389742 left12:08
← tetha left12:08
← gaveen left12:08
→ a389742 joined12:08
→ tetha joined12:08
← rchavik left12:09
← s1 left12:09
← Blub\0 left12:09
← levifig left12:09
← Borbus left12:09
← Kudos left12:09
← beppu left12:09
← ewanm89_ left12:09
← w-m- left12:09
← Derega left12:09
← tank__ left12:09
→ rchavik joined12:09
→ Blub\0 joined12:09
→ w-m- joined12:09
→ tank__ joined12:09
→ s1 joined12:09
→ beppu joined12:09
→ Derega joined12:09
→ ewanm89_ joined12:09
→ levifig joined12:09
→ Borbus joined12:09
→ Kudos joined12:09
lamont` → lamont212:12
lamont2 → lamont12:13
← alireza__ left12:13
→ selckin joined12:13
→ maleknet joined12:14
reuss is there no way of looking at refs outside of refs/heads, from a remote repository? .. or do i have to know the sha1? ..12:15
sitaram git ls-remote12:16
should be under refs/ though, but need not be refs/heads or refs/tags12:17
and even if you know the SHA1 it wont help you; nothing in git will let you grab a SHA unless it has a ref pointing to it12:17
(security reasons, I guess)12:17
reuss oh yeah -- then it only worked because i had the same sha1 in my own version, which makes sense..12:18
→ yaotti_aw joined12:18
yaotti_aw → yaotti12:18
← johan-s left12:18
reuss well, other question then - is it possible to reference it by the "refs/*" name in my local repository?12:18
→ nachocab joined12:18
nachocab how do you make Paperclip change the location of a previously uploaded file? (assuming I changed the has_attachment definition)12:20
← kpreid_ left12:20
nachocab oops, my bad12:20
← nachocab left12:20
sitaram reuss: didn't understand the question :(12:21
→ qrush_ joined12:21
qrush → qrush|away12:22
reuss sitaram: you can do: git log origin/<branch> .. is there a way of doing what with refs stored in say refs/custom ? ...12:22
erh, what=that12:22
sitaram refs/custom on your local side?12:22
← gk__ left12:23
→ NET||abuse joined12:24
dominikh just reading about git attributes and .gitattributes/.git/info/attributes. is there also a way for specifying global attributes that apply to all of my git repos?12:26
→ gk joined12:26
reuss sitaram: yeah - so you don't need to do git ls-remote origin - to see where it's stored .. or are they considered private?12:26
→ psynaptic joined12:27
engla dominikh: no, you can't12:27
← _jason553839 left12:27
dominikh too bad12:28
i liked the idea of being able to diff binary files using external tools like "strings"12:28
doener reuss: try "git config --get remote.origin.fetch"12:29
reuss: that shows the default fetch refspec for that remote12:29
→ _jason553839_ joined12:30
doener reuss: by default, "git clone/remote add" sets that to "+refs/heads/*:refs/remotes/origin/*" (for a remote called "origin")12:30
reuss doener: oh so you should configure it to also fetch what other refs you're insterested in?12:30
→ jas4711 joined12:30
→ Fullmoon joined12:30
← pom__ left12:30
doener reuss: so "fetch" only fetches things from the refs/heads/ namespace, if you want to fetch other stuff, you'll have to do that manually (specifiy a refspec on the command line), or adjust the config12:30
reuss: you can have more than one refspec in the config, just add more "fetch = ..." lines in the file12:32
reuss: or use "git config --add remote.<name>.fetch ..." (the --add makes it add another entry, instead of overwriting the existing one)12:32
reuss oh, cool. :)12:32
← pom_ left12:33
→ pom joined12:34
← jinzhu left12:36
→ tstclair joined12:38
→ khmarbaise joined12:38
→ wolog joined12:39
→ _jason553839 joined12:39
→ _ringlej joined12:41
sitaram doener: do you know of a way to make a 3-way merge disregard indentation differences? Causing too many conflicts :(12:41
doener: or disregard *all* whitespace, like diff -b12:42
→ bdimcheff joined12:42
← bdimcheff left12:43
→ bdimcheff joined12:43
doener sitaram: even -b doesn't really "drop" the whitespace changes, it just disregards lines that have whitespace changes _only_12:43
← bdimcheff left12:44
doener sitaram: how many files are we talking about?12:44
sitaram doener: not my project, but I think about 40-6012:44
doener sitaram: hm, I can think of some really really ugly thing one could do...12:45
sitaram preprocess them and use that?12:45
doener sitaram: but I'm not even remotely sure that it would actually help12:45
sitaram: kind of12:45
sitaram tell me anyway; even if it for academic curiosity I want to know12:45
← _jason553839_ left12:46
→ DavidKlein joined12:46
doener for each unmerged file that has modify/modify conflicts (i.e. stages 1-3 all exist in the index):12:46
grab the common ancestor, diff -b against the side on which you want to ignore lines that only had whitespace changes, apply the patch12:47
sitaram aah -- should work; I can only try it tomorrow, but i will12:47
let me think. diff -b :1: :2: and apply that patch to :3:?12:48
doener: ^12:48
doener then use either "git merge-file" (needs the files in the working tree, so get them there), or use "git merge-one-file" (needs a blob, so run "git hash-object -w" first)12:48
→ hugoxrosa joined12:50
doener git show :1:path/to/file > path/to/file; git diff :1:path/to/file :3:path/to/file | git apply; git merge-one-file :1:path/to/file :2:path/to/file $(git hash-object -w path/to/file) path/to/file12:50
sitaram: like that, I think12:50
sitaram doener: thanks; I see that merge-one-file has a rather sparse manpage, no hint about what arguments it takes :(12:51
doener sitaram: it's a shell script, arguments documented in there12:51
sitaram doener: ok12:51
← tjboudreaux left12:51
sitaram doener: care to spend a couple minutes explaining that to me; just at a high level?12:52
doener sitaram: the basic merging internally does something like:12:52
← galderz left12:52
→ jason553839 joined12:52
→ roop joined12:52
doener sitaram: git merge-one-file :1:file :2:file :3:file file [...mode stuff...]12:53
sitaram: just the plain 3-way merge12:53
→ johan-s joined12:53
doener sitaram: the show + diff + apply thing creates a new "theirs" version of the file, which doesn't have the whitespace-only changes12:54
sitaram doener: ok...12:54
aah12:54
doener sitaram: show :3:path > path; diff :1: :3: | apply ==> you get :3: again12:54
sitaram doener: didnt you mean diff -b there?12:54
doener sitaram: with the -b in there, you get :3: without the whitespace stuff12:54
sitaram oh12:55
doener sitaram: oh, oops, yeah12:55
sitaram scratches his head12:55
doener sitaram: the "long hack" should have "diff -b"12:55
sitaram: otherwise, as I just said, we simply re-create the :3: version12:55
sitaram ok12:55
doener sitaram: and the final "git merge-one-file" call just does the merge with the new version we just created. The "git hash object -w" thing just puts it into the object db, so merge-one-file can use it12:56
sitaram doener: and the first 2 args dont need that because they already exist (the :1:file syntax *is* a blob sha) right?12:57
doener sitaram: right12:57
sitaram doener: up here: 18:24:37 < doener> sitaram: show :3:path > path; diff :1: :3: | apply ==> you get :3: again12:57
doener: you meant show :1:? or is it really :3:?12:58
doener sitaram: I meant :1:12:58
doener fails12:58
doener sitaram: as I said, the hack is butt-ugly. And (if at all), I'd only use it if I know that I can recreate the right indentation/whitespace after having solved the (now hopefully easier) conflicts12:58
→ Pupeno joined12:59
doener sitaram: otherwise the merge result is some weird mess12:59
sitaram :012:59
→ bdimcheff joined12:59
sitaram doener: I have a feeling this might actually work for some people though; I'll let you know tomorrow sometime :)12:59
doener wonders how to interpret that smiley12:59
doener sitaram: I'm sure a lot of crap "works" for a lot of people. At least until they realize that it didn't13:00
→ therrg joined13:00
sitaram :0? interpret as bad keyboard/typing/shift key didn't "take" :)13:00
doener heh :-)13:00
sitaram doener: also as someone who's itching to go home before the traffic gets real bad :)13:00
→ flazz joined13:03
sitaram oh the "tomorrow" bit is also that it isn't my repo or I would have tried it right away :)13:07
← _jason553839 left13:08
← GodEater left13:08
sitaram doener: one last question: when you do merge-one-file A B C D, the 4th one is ignored/overwritten/immaterial what its starting contents are?13:08
doener sitaram: D is just the output file AFAIK, it gets overwritten13:09
← jason553839 left13:09
sitaram ok13:09
→ queso_ joined13:10
→ Bass10 joined13:10
→ kpreid_ joined13:12
← queso__ left13:13
→ chamo_ joined13:15
← khmarbaise left13:15
← chamo_ left13:16
→ flazz_ joined13:18
rchavik → spoonfeeder13:19
← gk left13:19
spoonfeeder → rchavik13:19
→ dsch04 joined13:20
← maleknet left13:22
← flazz left13:23
→ slabbeh joined13:24
→ abms1116 joined13:24
→ squentin joined13:25
→ fhr joined13:26
→ ryanarn joined13:26
→ fualo__ joined13:28
← hummingkid left13:30
→ tfarina joined13:30
← agile left13:32
← fualo_ left13:33
← chamo left13:33
← fhr left13:35
← sprasanna left13:39
→ TeckniX joined13:39
→ kukks joined13:40
→ Mezner joined13:41
→ ctusar joined13:44
→ ophswy joined13:44
ryah_away → ryah13:45
← flazz_ left13:46
→ cannonball joined13:47
→ drewolson joined13:49
→ suman joined13:50
→ roop_ joined13:52
→ [t0rc] joined13:55
→ mw joined13:57
← kpreid_ left13:58
→ flazz joined13:59
→ keystr0k joined13:59
← flazz left14:00
→ flazz joined14:00
→ flazz_ joined14:01
→ veiz joined14:04
→ hummingkid joined14:06
← roop left14:06
→ tschundeee joined14:07
← [t0rc] left14:08
→ makkalot joined14:10
← ophswy left14:11
makkalot hi how can i pull everything up to a given tag,for example i want to pull my v1.0 of my application ?14:11
→ nighthwk1 joined14:11
Ilari makkalot: pull as in fetch + merge?14:12
← rchavik left14:13
→ spearce joined14:13
← alezandro left14:13
→ Swimming_Bird joined14:14
→ nsussman joined14:15
→ akahn joined14:16
← spearce left14:16
sitaram Ilari: just FYI; gitolite will not be putting out that "need SSH_ORIGINAL_COMMAND" error; see http://github.com/sitaramc/gitolite/blob/pu/doc/3-faq-tips-etc.mkd#myrights for what I do with that now! [still in "pu", not yet in "master" though]14:16
→ spearce joined14:16
→ altrux joined14:17
← flazz left14:17
→ loincloth joined14:19
→ Turias joined14:19
sitaram Ilari: I hope you don't mind me sending you this sort of stuff apropos of nothing; I do it because you spend more time than all the rest put together helping people with gitosis/gitolite type questions and I want to make sure you're updated :)14:20
→ ph^ joined14:21
Ilari sitaram: Ok.14:21
→ bcardarella joined14:23
→ lucsky joined14:24
← ph^ left14:24
→ patcito joined14:26
← Swimming_Bird left14:26
→ gk joined14:27
→ icwiener joined14:29
flazz_ → flazz14:36
← Teo` left14:37
→ agile joined14:39
← ntoll left14:40
→ joevandyk_ joined14:41
→ drubies joined14:41
joevandyk_ how can i get a list of all branches that were branched off a given branch?14:41
→ queso__ joined14:41
← kent\n left14:43
→ kent\n joined14:43
Ilari joevandyk_: Git doesn't have hierarchial branches...14:44
→ chamo joined14:45
jacobat joevandyk_: You might be able to use "git branch --contains [commit]"14:45
chamo sitaram: are you still around?14:45
jacobat It's not quite the same though14:45
← Lisimba left14:48
→ rolfb joined14:49
→ argnel joined14:49
← cannonball left14:50
← argonel left14:50
← __bert__ left14:50
← cj left14:50
→ jstemmer joined14:53
← queso_ left14:54
rolfb → _rolfb14:55
→ ACSpike[Work] joined14:56
→ alezandro joined14:56
→ Haarg joined14:58
sitaram chamo: yes but cant talk right now; maybe another 20 min?14:59
sorry :(14:59
chamo sitaram: np. okay. cause my passwordless access suddenly drop and not working after ssh'ing to other hosts.14:59
ACSpike[Work] a while back I cloned a repository that has ~325 branches, but I'm only interested in one or two. when I `git fetch` all of them get fetched and one of them seems to be causing an error (missing object somewhere). can I delete one or some or most of the files from inside .git/refs/remotes/origin/. could that possibly solve my problems? (waiting for lots of code I'm not interested in and having an error with one particul15:01
Ilari chamo: ssh agent?15:01
chamo Ilari: I don't know...15:01
Ilari ACSpike[Work]: Missing object errors are fairly serious...15:02
offby1 ACSpike[Work]: you can "git branch -D"15:03
→ cannonball joined15:03
chamo Ilari: what can I check on ssh agent?15:03
Ilari chamo: What key it tries... No idea how to determine that...15:04
chamo Ilari: ssh -v user@host15:04
Ilari chamo: You could try unsetting 'SSH_AUTH_SOCK' and then trying access gitolite repos...15:04
chamo the verbose option answers this15:04
Ilari chamo: And define not working... User info from gitolite or not a git repository error?15:05
→ madewokherd joined15:05
ACSpike[Work] Ilari: the error http://gist.github.com/22052815:06
chamo Ilari: it's the ssh priv/pub key that suddenly doesnt work. it was working before.15:06
Ilari: and I didnt change any options15:06
I did just login on other ssh servers15:06
← taicomjp left15:06
Ilari chamo: Or passwrd prompt?15:06
chamo with regular login/password15:06
it shows me a pass prompt15:06
← WebDragon left15:06
chamo the pub/priv key is not found anymore on local comp15:06
but all is OKAY15:06
← veiz left15:06
chamo as it hasnt changed15:06
...15:06
→ sagsousuke joined15:06
→ WebDragon joined15:07
Ilari ACSpike[Work]: Can you try using git://?15:07
chamo: With SSH_AUTH_SOCK unset and ssh with -v, does it show the right key tried?15:07
chamo: (might require more -v 's for it to show what it actually tries).15:08
chamo why unset SSH_AUTH_SOCK ?15:08
Ilari chamo: So ssh-agent won't come messing things up.15:08
← tvw left15:09
ACSpike[Work] Ilari: I'm limited by my ignorance of git. I could open git:// through the firewall, but I wouldn't know how to use git:// instead of http:// at this stage in the game. I'll reasearch it if you think it could help. but I'm also happy to stop fetching all these branches.15:09
← qrush_ left15:09
Ilari chamo: Always good to test with minimal configurations and at low level and then add parts until reaching real scenario.15:09
→ qrush joined15:12
Ilari ACSpike[Work]: 9418 outbound is not open in firewall? git:// is nicely behaved protocol (single TCP connection client to server).15:12
chamo Ilari: ...15:12
okay15:12
→ Swimming_Bird joined15:12
Ilari ACSpike[Work]: I think that you could change the URL in configuration file (to git:// URL of repo) and then retry the fetch...15:13
chamo Ilari: not working15:14
← drewolson left15:14
Ilari chamo: Not trying right key?15:15
ACSpike[Work]: I guess the right URL would be git://git.catalyst.net.nz/moodle-r2.git ... I ls-remoted it, it does exist...15:16
chamo Ilari: okay i've tried what you said, still asking for password.15:17
→ dunolie_ joined15:17
chamo then I exported the SSH_AUTH_SOCK again15:17
Ilari chamo: In verbose ssh output, can you confirm it is trying right private key?15:17
chamo and I cleaned with ssh-add -D15:17
and added again my 2 priv keys15:17
tried to log again without success15:18
Ilari chamo: Right now, what needs to be determined is wheither login failure is because client doesn't send the right key or because the server rejects it.15:18
chamo okay15:19
← therrg left15:19
chamo well, the client is offering the different private keys15:19
but then nothing happens15:20
sitaram Ilari: I have a feeling something is changing the authkeys file on the other side...15:20
Ilari chamo: And at least without ssh-agent, with sufficient number of -v's, the debug output of ssh will tell what keys are tried and in what order (the first key must be correct).15:20
chamo and the last auth method is password15:20
Ilari: it offers the wrong first!15:20
Ilari chamo: With gitosis/gitolite: password -> failed.15:20
sitaram chamo do you have gitosis also setup by any chance? or something else that may be fouling up the remote side authkeys file?15:20
← elTigre left15:21
← fualo__ left15:21
ACSpike[Work] Ilari: thanks. I was just about to report failure :-)15:21
sitaram chamo: or someone else (horrors!) also has authority on that account and is doing stuff?15:21
ACSpike[Work] but it seems to be doing something now with the fixed url15:21
chamo sitaram: yes all set up. it worked before. like what we looked earlier. but now, for a strange reason, it doesnt15:21
Ilari: yes with gitosis it fails.15:22
sitaram and this is the second time it has done this? (works, suddenly stops working)?15:22
chamo seems so15:22
Ilari: but what i dont understand is that it was working well 2 hours ago15:22
and I worked on other project, didnt touch the git server at all15:22
Ilari chamo: Is that "wrong" key it tries supposed to have some kind of access (either full-power shell or gitolite key)?15:22
chamo just used the client to ssh to another remote server15:22
on which we use the "classic" l/p auth method15:23
← johan-s left15:23
→ johan-s joined15:23
sitaram Ilari: agent forwarding screwing up something?15:23
chamo gitolite, not gitosis15:23
→ koltroll joined15:23
→ patmaddox joined15:24
Altreus How do I point a branch name at a different commit?15:24
koltroll hey guys. I was here some weeks ago trying to get git push web to work by adding post-receive hooks etc. The problem then was that checkout -f failed because I had an old version of git on the server. Now I have updated git, tho I'm still having troubles. but now I'm not getting -any- errors, so I guess that's good. But I'm also not getting any results.15:24
doing git push web pushes the content to right repo, but the post-receive hook doesn't seem to work.15:25
Ilari Altreus: update-ref for local refs not checked out, for checked out branch, reset, for remote branches, push.15:25
koltroll the post-receive hooks more or less only contains of checkout -f, also I've chmod +x on it.15:25
→ elmob joined15:25
Altreus fanks15:25
koltroll uh. I hadn't done +x properly.15:26
→ Lisimba joined15:26
chamo Ilari: well when I remove the env var, with verbose SSH login, it doesnt use the right priv/pub keypair15:26
koltroll HOly FUCK15:26
IT WORKS15:26
chamo lol15:27
Ilari chamo: What really matters is first keypair it tries that is supposed to have access (or wheither it tries such key at all).15:27
→ Theravadan joined15:28
→ fualo_ joined15:28
Ilari chamo: Note that if you have dual key setup, you have seperate host alias for accessing the server using gitolite key.15:29
chamo Ilari: yes I understood that. Well i tried to remove ALL keys, then add only the one I need15:29
→ mw___ joined15:29
chamo yes exactly!15:29
← sdboyer left15:29
← niki left15:29
chamo damn15:29
→ rchavik joined15:29
sitaram chamo: http://pastebin.com/m788481015:30
→ fhr joined15:30
→ sdboyer joined15:30
→ warthog9 joined15:30
sitaram that's a little more complex than it needs to be because its a cut from something else but it should work15:31
→ javatexan joined15:31
← dunolie left15:31
chamo sitaram: mmmh15:31
Altreus I did $(git update-ref branch SHA1) and now it says branch is ambiguous15:31
chamo sitaram: why change something on server15:31
I havent touched it15:32
← jaysern left15:32
chamo okay lets try15:32
sitaram chamo: we're not changing anything in that script; we're just getting all the key fingerprints15:32
tvon /join fab15:32
Altreus oh I should try this -d15:32
→ jaysern joined15:32
chamo okay15:33
← mw left15:34
mw___ → mw15:34
→ smuf joined15:34
Altreus hm15:34
what does it mean when it says a refname is ambiguous?15:35
← sdboyer left15:35
→ sdboyer joined15:36
→ hobodave joined15:37
→ elTigre joined15:38
← wolog left15:38
Altreus is typing in invisible ink15:38
→ taicomjp joined15:39
→ jofrum joined15:40
← _rolfb left15:40
Ilari Altreus: Multiple matches among the ways it tries to qualify ref.15:42
Altreus aha15:43
When I check out $branch it points me at $SHA1, but I used update-ref to point it at $SHA215:43
Ilari Altreus: You missed the refs/heads/ prefix?15:43
Altreus ah15:43
I'll try this15:43
← spearce left15:43
Ilari Altreus: I think update-ref can also delete refs.15:43
← sdboyer left15:44
Altreus it can - but it told me it wasn't part of it. the /refs/heads seems to have done it15:44
→ sdboyer joined15:44
Altreus er refs/heads15:44
jpic_ hi, what does that mean please? http://nopaste.com/p/aO4hh4vvC15:45
trying to clone a bare repo15:45
← chamo left15:45
← sdboyer left15:46
jpic_ my bad, i think i failed at reading the manual, sorry15:46
→ sdboyer joined15:47
← jpic_ left15:47
← hgimenez left15:47
→ qhoxie_ joined15:48
← __iron left15:56
← fhr left15:57
→ fhr joined15:57
→ patrikf joined15:57
→ babilen joined15:58
← Weasel[DK] left15:59
→ Weasel[DK] joined15:59
→ jrmuizel joined15:59
→ juan_arandaalvar joined15:59
← sde_antrix1 left16:00
→ GodEater joined16:00
← efaistos left16:01
→ futoase joined16:01
← Pupeno left16:01
→ spearce joined16:01
→ Pupeno joined16:02
← Pewpewarrows left16:02
← dancor left16:04
← abms1116 left16:04
← johan-s left16:04
→ Pewpewarrows joined16:05
← hegge left16:07
→ JasonWoof joined16:10
← jrmuizel left16:11
→ btanaka joined16:12
← a389742 left16:13
← wagle left16:13
← fualo_ left16:13
→ fualo_ joined16:14
← Azeroth-Work left16:17
← psoo left16:17
→ agib joined16:17
← fualo_ left16:17
← thaz left16:17
→ JoeConyers joined16:17
← fluxusps left16:17
← s0ber left16:20
→ jmcantrell joined16:21
→ priidu joined16:21
← ACSpike[Work] left16:23
← tvon left16:25
← WALoeIII left16:26
← bentob0x left16:27
← JohnFlux left16:27
argnel → argonel16:30
→ dreiss joined16:30
← slabbeh left16:31
→ parasti joined16:34
→ galderz joined16:35
→ yang_ joined16:40
→ NesWork joined16:40
NesWork hi room.. need some help on git here..16:41
← futoase left16:41
→ WALoeIII joined16:42
NesWork it says " warning: too many files, skipping inexact rename detection".. :-(16:42
← hingwah left16:42
← drubies left16:42
Ilari NesWork: Increase the limit or live without inexact rename detection...16:44
← argonel left16:44
→ Teo` joined16:45
← aatanasov left16:46
→ rettub joined16:46
Ilari NesWork: config setting diff.renameLimit16:46
→ lhz joined16:48
→ Axius joined16:48
→ argonel joined16:48
→ vuf joined16:48
NesWork Ilari, where can i find the conf file?16:49
→ marcelo__ joined16:49
marcelo__ hello16:49
Ilari NesWork: man git-config16:50
Gitbot NesWork: the git-config manpage can be found at http://git.or.cz/man/git-config16:50
marcelo__ what is the difference of using git revert HEAD and git reset HEAD^ for rolling back a commit16:50
?16:50
teuf git revert adds a new commit which cancels the changes made in the latest commit16:50
and git reset just remove the commit16:50
Ilari marcelo__: Git revert HEAD is to reverse-cherry-pick HEAD. 'git reset HEAD^' resets branch and staging area to last commit (but leaves working copy alone).16:50
NesWork Ilari, Gitbot thx. :-)16:51
← qhoxie_ left16:51
teuf ie in the 1st case, you'll have 2 commits in git log, in the 2nd case, no commits16:51
mee so revert will also remove the change associated with the most recent commit from your working copy?16:51
that's a pretty big difference, too16:51
→ qhoxie_ joined16:52
marcelo__ Ilari, thank you16:52
← mattayers left16:52
← Teo` left16:53
→ segher joined16:53
← mw left16:54
→ Teo` joined16:56
→ abms1116 joined16:57
← GarethAdams|Work left16:57
← Thumper_ left16:58
← babilen left16:58
→ wagle_home joined17:00
→ hegge joined17:01
← jmcantrell left17:02
→ SRabbelier joined17:03
← flazz left17:05
→ flazz joined17:05
← froschi left17:05
← warthog9 left17:06
→ tansell_laptop joined17:09
→ khmarbaise joined17:10
→ ashc joined17:11
→ rudi_s joined17:14
→ veiz joined17:14
← murph_ left17:15
← galderz left17:16
→ murph_ joined17:16
→ mjansen joined17:17
← veiz left17:17
→ mrlemao joined17:18
mjansen any idea why "git blame <full_path>" doesn't work? It seems like git tries to find a repository from the current directory instead from the target file17:18
i'm in the build directory trying to issue that for a file from the source directory17:19
→ krawek joined17:19
→ jsnmap joined17:20
→ veiz joined17:20
→ silas_ joined17:22
← mrlemao left17:23
dbt mjansen: easy enough to fix; (cd `dirname $file` && git blame `basename $file`)17:24
mjansen dbt: i know. i just wonder why it doesn't work out of the box?17:24
it's not only that command. all of them17:24
→ bdiego joined17:25
→ mbroeker joined17:26
→ khmarbaise_ joined17:26
← Weasel[DK] left17:26
→ khmarbaise__ joined17:26
→ Diapo joined17:26
← khmarbaise left17:27
→ Weasel[DK] joined17:27
→ tbuehlmann joined17:27
khmarbaise__ → khmarbaise17:27
← Grahack left17:27
→ icwiener_ joined17:29
← cedricv left17:29
← saurabhverma left17:31
← icwiener left17:31
← Axius left17:31
→ lohrun joined17:32
← jsnmap left17:32
→ Sigma joined17:33
ashc sitaram: are all the hooks working in latest gitolite master?17:33
Yewbacca Holy cow http://digg.com/politics/Arnold_Schwarzenegger_to_Legislators_F_You_No_really17:33
→ kennethlove joined17:34
→ AlexB joined17:35
← silas left17:35
← holizz_ left17:36
levifig does removing a remote branch remove it locally as well?17:36
← Fullmoon left17:37
← shruggar left17:40
← mcella left17:40
← veiz left17:40
→ bugfux joined17:41
← abms1116 left17:42
→ oosp joined17:44
→ dirk2 joined17:44
← khmarbaise_ left17:44
→ babilen joined17:45
steveire → steveire_17:46
→ rylar joined17:46
→ drubies joined17:46
oosp hello! i hope this isn't a silly question (i couldn't find an answer via Google) but is there an easy way to show the submodule version in a tree-ish? specifically I want to do something like: "git show origin/master:name-of-submodule" and have that give me the version of the submodule in origin/master17:47
Ilari levifig: 'git remote prune <remote>'.17:47
oosp: 'git rev-parse origin/master:name-of-submodule'?17:47
levifig Ilari, what does that do?17:47
I want to remove it remotely but NOT locally17:48
Ilari levifig: Delete remote tracking branches that no longer exist on remote.17:48
← priidu left17:49
Ilari levifig: If you want to leave the local branch alone, then just leave it alone. Pushing deletion to another repository won't affect local branches in current repo (unless you specify current repo as target for push).17:49
oosp <Ilari> oosp: 'git rev-parse origin/master:name-of-submodule'? <-- perfect, thank you :) i didn't realise git rev-parse knew that syntax too17:49
Ilari oosp: Its generic sha-expression syntax.17:49
← Renegade_ left17:49
← khmarbaise left17:50
oosp Ilari: ah, OK - thanks again17:50
← zoniq left17:51
← elmob left17:51
← Tuomas left17:52
WebDragon if I've ever done a git commit --amend is the intermediary commit still somewhere in the history ?17:52
← corni left17:54
→ dmq joined17:54
→ warthog9 joined17:54
dbt WebDragon: try the reflog (git log -g or git reflog)17:54
WebDragon tips hat17:55
→ godsflaw joined17:56
← rchavik left17:56
→ Adlai joined17:57
WebDragon → WebDragon|away17:58
← struberg left18:04
→ jrmuizel joined18:04
→ psoo joined18:05
→ tg_ joined18:07
→ alsadi joined18:09
→ ph^ joined18:12
→ mib_56hh88 joined18:12
tk hmm maybe someone has an answer to this... I am following http://toroid.org/ams/git-website-howto which sitaram pointed me at yesterday... however my git repo has files that should never be published (like a resources used to make images and what not) -- is it possible to only have certain portions of the site checkout to the live web dir?18:13
← mib_56hh88 left18:13
← agib left18:13
WebDragon|away → WebDragon18:13
alsadi I'm using cgit and I want to add a custom css class to the <pre> used in rendered HTML code ?18:15
is there something got that in cgitrc ? or any other way ?18:15
→ thiago_home joined18:16
→ veiz joined18:17
→ struberg joined18:18
→ kpreid_ joined18:19
alsadi s/got/for/18:20
→ ihoka joined18:20
→ justin-george joined18:20
→ richcollins joined18:21
Mizar-jp gitk Japanese Translation 暫定的ですが日本語訳が完了しました。gitkを利用している日本語使用者の方は、おかしな訳が残っていないかレビューをお願いいたします。 http://lab.mzr.jp/gitk/18:24
← richcollins left18:25
← bugfux left18:26
→ tono_ joined18:26
→ agib joined18:27
Mizar-jp (Japanese translation is completed preliminary. If you are using gitk and Japanese, please review whether there is a strange translation.)18:27
← rtyler left18:27
mugwump you mean "completed preliminarily" and "whether there are any strange translations" ? ;)18:29
→ nirvikalpa joined18:30
Mizar-jp I'm sorry, I do not speak English because it is using a machine translation.18:31
← cilly left18:31
→ rtyler joined18:31
→ bdimcheff_ joined18:32
→ cilly joined18:33
→ mm_work joined18:33
→ Axius joined18:33
← bdimcheff left18:33
bdimcheff_ → bdimcheff18:33
← ph^ left18:34
Mizar-jp What phrase was hard to understand?18:35
→ jmcantrell joined18:35
mugwump I'm just trolling you18:35
well done on completing your translation :)18:36
← tono left18:37
→ tono__ joined18:37
← warthog9 left18:39
← veiz left18:40
Mizar-jp Hmm, I do not know what it was intended to do, but thank you.18:41
→ Arrowmaster` joined18:44
→ richcollins joined18:44
← Arrowmaster left18:45
Arrowmaster` → Arrowmaster18:46
← juan_arandaalvar left18:46
→ khmarbaise joined18:46
← tono__ left18:48
Mizar-jp "The contents of the Japanese of the previous line, it is said in English?" And wanted to say what? That's right, if that.18:50
← roop_ left18:51
godsflaw hey guys, I _need_ a clone or an archive to preserve timestamps on files. Does anyone have ideas on how to make this work?18:52
← Weasel[DK] left18:53
→ Weasel[DK] joined18:53
← tono_ left18:53
Ilari godsflaw: Store it in seperate files and restore on checkout?18:54
Adlai godsflaw: You could theoreticall add hooks to store and restore timestamps...18:54
→ Dashkal joined18:54
godsflaw right now the only option we can see is keeping a checked out master and building packages off that.18:55
→ coherence joined18:55
bremner godsflaw: use git archive?18:56
mugwump Mizar-jp: I was making a joke out of correcting your English, see this for more info: http://www.youtube.com/watch?v=XbI-fDzUJXI18:56
coherence any gitosis users - i'm trying to get the post-receive hook to work for commit emails working. any hints?18:57
godsflaw bremner: we are now and it does not preserve timestamps.18:57
bremner use the right options for tar?18:58
godsflaw we tried, but those need to be passed on the compress side, and it does not look like we have a solution for that.18:58
bremner or yeah, there is some fixup the metadata tools around. I forget the names.18:59
nirvikalpa ponnies == trojans?18:59
bremner godsflaw: metastore is one apparently19:02
http://david.hardeman.nu/software.php19:02
WebDragon → WebDragon|away19:03
Mizar-jp mugwump: Unfortunately, your jokes and jokes that video, I do not understand much.19:03
cannonball Am I seeing this correctly? git rm path/to/dir/file will drop that file. If it was the only file in that directory, it apparently drops the now empty directory from the repo as well. Intended? Or unintentially did exactly what I wanted? :-)19:05
← bcardarella left19:05
← qrush left19:07
mugwump Mizar-jp: jokes are often "lost in translation"19:08
Ilari cannonball: Intended.19:09
← smuf left19:10
Adlai cannonball: Git only tracks files, not directories. If you had wanted to track the directory, you could put an empty .gitignore file in it.19:10
← drubies left19:13
→ ramus`_ joined19:15
← Axius left19:15
← richcollins left19:18
← head left19:19
→ Dashkal_ joined19:19
→ doy joined19:20
← Dashkal left19:21
Dashkal_ → Dashkal19:21
doy is there a way to pass default command line options to git commands?19:21
← bdiego left19:21
doy i'd like "git log" to always be "git log --stat"19:21
Ilari doy: No.19:21
← mjansen left19:21
doy bah19:21
any workarounds?19:22
mugwump doy: git config --global alias.logs 'log --stat'19:22
then 'git logs' will do whbat you wanted19:22
nirvikalpa `alias gl='git log --stat'19:22
doy sure, i'd rather have it on just log though19:22
oh well19:22
← tansell_laptop left19:23
← btanaka left19:25
→ __iron joined19:25
→ benn_d joined19:25
← benn_d left19:26
mugwump not sure why you'd care, but ok19:26
probably a config option for that would be vetoed due to performance reasons19:26
→ ChrisGo joined19:26
mugwump and because you can use alias19:26
doy performance?19:26
mugwump sure, well, you're having to diff everything to get the --stat19:27
doy oh, for --stat specifically, yeah19:27
ChrisGo is git as nice as the collab svn for windows server installer - ease of use is important for me right now19:28
→ richcollins joined19:29
← capndiesel left19:29
→ peff joined19:29
engla does anyone have a nice tutorial for topic branch workflows? That I can send to collaborators, without having to write it myself19:29
jonthefly ChrisGo: There are alot of good things i could say about git... but i dont think i would describe it as easy19:30
especially if your new to the git concepts19:30
← ramus` left19:30
ramus`_ → ramus`19:30
← doy left19:30
mugwump the only thing that's easy about svn is it's mum19:30
jonthefly svn might be a better choice if you do not care about branching and merging.19:30
ChrisGo well the gui for windows looks just like the svn version19:31
just making sure its easy to install as server on windows without a whole bunch of non native crap19:31
mugwump instead you have to install native crap :)19:31
ChrisGo updates, and such... collab has done a pretty good job with this19:32
nirvikalpa There was a tutorial for people who don't know distributed scm..19:32
mugwump ChrisGo: you have to install mSysGit, it has a couple of dependencies, but most people manage it ok19:32
whether that's easy or not, well, that's always up to the material you're working with19:32
ChrisGo looking at msysgit19:33
← nirvikalpa left19:33
ChrisGo cygwin is a pain imo19:33
← dpino left19:34
mugwump msys != cygwin19:34
the only reason it's msys and not plain mingw is for crappy svn api19:35
Thell Got a four part msg, sorry for the length...19:35
ChrisGo right thats why I was looking at that no native C build for windows yet?19:35
Thell ooops, thought you two where at a breaking point there19:36
Thell will wait19:36
ChrisGo im done19:36
Thell Hi guys. Here's an interesting conundrum... A build version info script for VS ( http://git.pastebin.com/m7e15aba5 ) with a quirk I'm not quite sure how to handle efficiently.19:36
The 'digital' file version is based on git describe output from a tag. So tag v1.0.0 + 1 commit would be 1,0,3,1 ( 3==release version ), this is fine except that if we have a maint branch on put a patch on there for tag v1.0.119:36
The digital version is then 1,0,3,1. See the problem? two different commits giving the same digital version. Even though the string FileVersion is different and can clarify the situation.19:36
Any ideas?19:36
mugwump Thell: use 'git describe'19:37
← ChrisGo left19:37
Thell mugwump: obviously I am19:37
...'based on git describe output'...19:38
→ ianmcorvidae|alt joined19:38
mugwump so, well, don't throw away the bit which makes it different19:38
Thell 'digital' version means no alpha19:38
mugwump the -gcafebab part of the version name19:38
Thell four part comma seperated int value19:39
mugwump sure, well don't do that then19:39
add the hash19:39
like git describe does19:39
bremner or use tags?19:39
Thell your not getting it... the digital version doesn't allow alpha mugwump19:39
mugwump well you're poked then19:40
→ tvw joined19:40
Thell bremner: tags are being used, that's how the 'git describe' is getting the # of commits since the tag19:40
bremner I mean, just tag the commit you mean to correspond with your version.19:42
Thell I did think briefly about using a '00', '01', etc... value at the end of the string for tagged release commit differences, so 1,0,3,0119:42
mugwump you could convert the hash to an int19:42
Thell bremner: we are19:42
mugwump actually for debian version numbers I tend to use a timestamp19:42
Thell actually I guess it doesn't have to be an int, just numeric19:42
I wanted to do timestamp too... I was outvoted for 'ease of reading' :/19:43
mugwump eg, 1.2.3.0.2009082119:43
→ wolog joined19:43
→ bcardarella joined19:43
← ianmcorvidae|alt left19:43
mugwump I think you might just have to say that functionality comes before astroturfing19:43
← voker57 left19:43
Thell ohhh and another issue with that was something to do with pdb info between different users for the same build...19:44
→ ianmcorvidae|alt joined19:44
mugwump not quite sure how a timestamp is not easy to read19:44
→ joevandyk__ joined19:45
mugwump you could split the time part19:45
1.2.3.0.20090821.12345619:45
mugwump helps19:45
→ Xaositect joined19:46
← ianmcorvidae left19:46
→ voker57 joined19:46
Thell four part19:47
← kpreid_ left19:48
→ ihoka_ joined19:50
mugwump sounds like you're really been run into a corner by stupid requirements19:50
← richcollins left19:51
→ richcollins joined19:52
← GodEater left19:52
→ bigmack83_ joined19:52
Thell I guess it really depends on the debug file info, from what I was told all file version info needed to match to be able to load a dump.19:52
→ GodEater joined19:52
← nsussman left19:52
← Xaositec2 left19:53
← tstclair left19:54
bigmack83_ ok i have two clones of a repo. on one i added a branch with: 'git branch development' and comitted and pushed changes. in other repo i ran git pull and it showed new branch of development. but when i try to do 'git checkout development' i get the error: error: pathspec 'development' did not match any file(s) known to git19:54
← yang_ left19:54
→ tstclair joined19:54
← richcollins left19:54
bigmack83_ do i have the process wrong to add branches so others can use it? beforei was able to commit the second branch i had to change a file to be able to commit.19:55
Ilari bigmack83_: Maybe it isn't local branch yet (git checkout -b development remote/development) or in newer git, shorter form (git checkout -t remote/development).19:55
← marcelo__ left19:55
Ilari bigmack83_: See 'git branch -r'.19:55
bigmack83_ well it was all done on the server. i hadnt tried over ssh yet.19:55
ok one sec.19:56
→ arrenlex joined19:56
bigmack83_ ok in the second the development branch is listed19:56
arrenlex Hi. I am resolving merge conflicts after git stash apply. How do I make git throw out my local changes for one file and use the upstream version?19:57
← ihoka left19:59
ihoka_ → ihoka19:59
→ khmarbaise_ joined20:00
→ kpreid_ joined20:00
gitster "git co HEAD that-file"?20:00
→ exygen joined20:00
gitster s/co/checkout/, that is.20:00
← javatexan left20:01
bigmack83_ well i want to switch to the whole branch20:01
not just one file20:01
← joevandyk_ left20:02
arrenlex gitster: Thanks!20:02
charon bigmack83_: in addition to what Ilari said, you mistook pull for fetch20:03
← koltroll left20:04
→ drizzd_ joined20:05
→ qrush joined20:06
→ ab3 joined20:06
→ juice_d joined20:07
→ corni joined20:08
← Pupeno left20:08
bigmack83_ charlls, ok so on the second clone i need to run git fetch instead20:10
→ Pupeno joined20:10
→ tjholowaychuk joined20:11
→ malumalu joined20:11
bigmack83_ charon, oops, last message was meant for you20:13
← elTigre left20:13
→ rudi_s_ joined20:15
← __iron left20:16
charon bigmack83_: indeed, and then make a local branch with a command like the ones Ilari gave. pull does a merge, so you may also want to see if running it resulted in history you didn't want to create20:16
(i.e., the branch you were on when you ran git-pull may contain extra commits)20:17
→ __iron joined20:17
→ rofflwaffls joined20:17
rofflwaffls hi, i keep getting a error: Unable to verify pack 4297ca0b1564361d528ebabaa35981e04412a1cf is available when trying to clone my repo20:18
USA_NO_1 if a git clone stalls over ssh can i resume?20:18
bigmack83_ ok so in first dir, i added the files, committed, and pushed to repo to have the base starting point. run 'git branch development' to add the branch, edit a file. commit, push20:18
rofflwaffls i'm cloning over http though20:18
ianmcorvidae|alt → ianmcorvidae20:19
bigmack83_ then i in second directory. cloned after first push. i would run git fetch ?20:19
← khmarbaise left20:19
khmarbaise_ → khmarbaise20:19
bigmack83_ to pull the development branch20:19
err20:19
then (git checkout -b development remote/development20:20
← akahn left20:20
bigmack83_ to checkout the dev branch. tight?20:20
charon bigmack83_: yes. note that by default, 'git push' does not push new branches, so the second push needs an explicit 'git push origin development'20:20
bigmack83_ yes20:20
rylar → rylar_off20:20
← Fissure left20:20
← drizzd left20:21
charon bigmack83_: also note that from there on you have three separate development branches (one per repo) and need to update/merge the local ones after fetching20:22
→ Fissure joined20:22
→ slabbeh joined20:22
dgs__ i'm trying to do a git merge origin, but am getting an error 'Entry '/config/locales/en.yml' not uptodate. Cannot merge". I know this means that my local copy has changes that would be overwritten, but i'm not sure how to make git let me merge the differences?20:24
→ elliottcable joined20:24
patrikf dgs__: if you don't need the local changes, git checkout /config/locales/en.yml20:25
rofflwaffls anyone? :(20:25
patrikf dgs__: if you do need them, git stash might be of use20:25
← Sho_ left20:26
dgs__ can't i get git to present a merged version that I can edit into shape?20:26
Phurl → ThatFnordingGuy20:26
→ tansell_laptop joined20:27
patrikf dgs__: git will do so if it can't merge automatically. but in any case the working tree needs to be clean20:27
i.e. have no uncommitted changes in affected files20:27
so either: throw them away (checkout), stash them away for later (stash), or commit them now (commit)20:28
← pavelz left20:28
← alsadi left20:30
← psoo left20:30
← hugoxrosa left20:30
← rudi_s left20:31
dgs__ hmm. crap. looks like i had forgotten to do a commit in there20:31
← ihoka left20:31
dgs__ had heaps of changes that were stopping it from working20:31
ta =)20:32
Thell coolio, I think the maint version building is figured out. There's a VC_FF_PRIVATEBUILD flag which I set if head is dirty. If two digital file-versions match the actual maint tag wouldnot have the flag set.20:33
→ nsussman joined20:33
Thell which also means the VS_FF_PATCHED is not set either.20:34
woohoo.20:34
boto → help20:36
help → boto20:36
boto → ehabkost20:37
bigmack83_ charon ok now, on the second i have 2 branches, master and remote/branch . i have 3 seperate repos because the one master, and each user would have a copy of development right?20:37
krynnotaur → krynnaway20:37
→ ^aristotle joined20:37
^aristotle is there a way to check out a file into the working copy but not into the index?20:37
from a previous version20:38
← cilly left20:38
bigmack83_ so if the second user edits files in development the first user would run a pull from the second user, that would merge changes assuming no conflicts20:38
← madewokherd left20:38
^aristotle or, vice versa20:38
→ ph^ joined20:39
dgs__ i'm getting a message that my branch and origin/master have diverged and have 7 and 6 commits each respectively. there's a few results for this on google, but nothing I can find that describes what it means / how to resolve it...20:39
← ph^ left20:39
bremner merge20:39
← ashc left20:39
bremner git merge origin/master20:39
^aristotle restating the question from scratch... can I say "put /some/silly/file from version X into the index but leave the working copy alone" somehow?20:40
→ Alien_Freak joined20:40
dgs__ bremner: was that for me?20:40
^aristotle dgs__: yes20:40
bremner dgs__: yes20:40
doener dgs__: "origin/master" is a remote tracking branch, basically a local mirror of the "master" branch head of "origin" remote20:41
Alien_Freak if I setup a project with a git submodule... ie. project A has a directory foobar which is clone of project B... whenever I checkout A, it'll also do a pull from B.. correct?20:41
doener dgs__: and your local "master" branch head has that setup as its upstream branch20:41
dgs__: what the message tells you is that there are 7 commits reachable from master that aren't reachable from origin/master, and 6 commits reachable from origin/master that aren't reachable from master20:42
^aristotle aha, I wanted `git reset treeish -- path`20:42
→ Aikawa joined20:42
^aristotle dgs__ might want to rebase instead of merging20:42
depending on the case20:42
dgs__ cool.20:43
← jds left20:43
doener dgs__: if you now do "git merge origin/master", that will add a new merge commit to "master", which make the commits reachable from origin/master also reachable from master20:43
→ jds joined20:43
doener dgs__: so the message will change, saying that you now have 8 commits (one more than before) reachable from master but not from origin/master, but no commits reachable from origin/master but not from master20:43
← pgalbraith left20:43
doener dgs__: i.e. the message is just for your information.20:44
← flazz left20:44
→ flazz joined20:44
doener dgs__: "usually" it means that you made 7 commits locally, but didn't push them yet, and that "origin/master" also saw new commits (which you haven't merged yet) since you branched20:44
dgs__ yeah, that would be about right - was trying to merge my code + my bosses with origin20:45
made a right balls up of it =)20:45
doener dgs__: if you messed with rebase (or other history rewriting stuff), the meaning is slightly different. Therefore, git uses the "neutral" form, and leaves the rest up to your interpretation, as you should know what you did ;-)20:46
← Aikawa_ left20:49
← merlin83 left20:50
← qhoxie_ left20:51
← NET||abuse left20:51
← lorandi left20:51
← bigmack83_ left20:52
→ ablemike joined20:53
← scientes_ left20:53
→ merlin83 joined20:54
← peff left20:55
← tjholowaychuk left20:56
→ priidu joined20:57
→ godsflaw_ joined21:01
← lucsky left21:02
→ ryoma joined21:05
← godsflaw left21:05
→ bremner_ joined21:06
→ Th0ma5 joined21:07
← keystr0k left21:08
← malumalu left21:08
← dirk2 left21:09
← schmukkk2_ left21:09
← guerrill1 left21:09
→ guerrilla joined21:09
← flazz left21:11
→ lresende joined21:11
← Aikawa left21:13
← mbroeker left21:16
→ madewokherd joined21:16
← elliottcable left21:17
← tg left21:22
tg_ → tg21:22
→ djszapi_ joined21:22
→ onigiri joined21:23
→ _patton joined21:23
← d0k left21:24
← djszapi_ left21:24
→ Aikawa joined21:24
← bremner_ left21:25
← ehabkost left21:27
← ctusar left21:28
→ sgh joined21:29
← bcardarella left21:30
← _ringlej left21:30
← DavidKlein left21:31
← RobertMeta_ left21:32
← mm_work left21:33
→ [M7] joined21:33
→ iron joined21:38
← jmcantrell left21:38
← __iron left21:38
→ RobertMeta joined21:39
← ixti left21:42
← babilen left21:42
→ jlilly joined21:43
→ matthewford joined21:43
← WALoeIII left21:44
godsflaw_ is there a quicker way than using git log to see a list of log dates for each file?21:45
→ marcelo__ joined21:48
thiago_home no21:48
← kent\n left21:50
→ kent\n joined21:51
← macmaN6789 left21:51
→ agenteo joined21:52
← tschundeee left21:53
→ schacon joined21:54
← Sigma left21:54
← iron left21:54
schacon spearce, you around?21:54
← Akufen left21:54
spearce yup21:54
schacon was the smart http technical doc patch rejected? i don't see it in pu21:55
or in the whats cooking list21:55
spearce held back. its not complete and doesn't fully match the code in pu21:55
schacon k21:55
→ WALoeIII joined21:56
mugwump schacon: http://article.gmane.org/gmane.comp.version-control.git/129734 ?21:56
schacon yup21:57
→ codejunky joined21:57
mugwump are you updating your rfc doc?21:57
schacon i was thinking of redoing it based off this example21:57
godsflaw_ → godsflaw21:58
mugwump cool, a lot of people want that21:58
spearce yea, that's a draft, see http://repo.or.cz/w/git/fastimport.git?a=blob;f=Documentation/technical/http-protocol.txt;hb=http-doc for a more recent draft maybe21:58
uh, ok, now the more recent draft is there21:59
→ ashc joined21:59
godsflaw okay, so let's say someone has touched a file (not tracked) in my repository and I want to reset --hard the repo. Is there a way to get it to blow that file away (without changing the timestamps on my working directory files?)22:01
Ilari Been thinking new rules for refspecs. More DWIM stuff, remove some really error-prone forms, add some new forms (to take advantage of some protocol features unexposed to UI in current versions).22:03
← bdimcheff left22:03
mugwump Ilari: examples or demoo doc patch?22:03
Ilari For destinations: qualify heads/, tags/ and remotes/ as one would expect. If fetch lines are available, fetch refs without destination according to them (otherwise these are errors).22:05
Pull would have single-part forms only.22:05
← jrmuizel left22:06
Ilari Push would get three-part forms: src:old: , :old:dest and src:old:dest These would control what gets sent as old field in update request. Force push is impiled for these.22:07
→ hgimenez joined22:08
← icwiener_ left22:08
← lhz left22:08
← loincloth left22:09
→ damo221 joined22:09
_patton We've been using a bare git repository on our webserver to push to with a post commit hook to update the dev/ site. Problem is, now we need to allow another user to update the bare repository, but the post commit hook updates dev/ with mucked up permissions. I've tried using su command in the hook, but its not being run correctly. Anyone?22:09
→ babilen joined22:09
damo221 i want to put a link on my website for people to apply a patch i wrote and build an experimental kernel module... but i want to specify the exact same git revision so the kernel will build successfully... how do i specify which git revision to check out?22:10
← gunmuc left22:10
thiago_home _patton: this is a shell scripting issue22:10
← Mezner left22:10
thiago_home _patton: not really a Git problem22:10
tk _patton: is your repo setup exactly how your site is pushed to the dev server? (working on this myself but my repo has non www files in it too)22:10
_patton thiago_home: yes true.22:11
tk _patton: maybe add your git user to the non pass required sudo list?22:11
_patton tk: our repo is set up with 3 branches, each for dev/ staging/ master/22:11
tk: tried that, hasn't worked22:12
Ilari damo221: Just say what kernel revision (by sha-1 hash or describe output) it is intended to be appiled on?22:14
→ sudoer joined22:14
sudoer i accidently added a 122 meg file to git, how can I completely remove it form git?22:15
← ablemike left22:15
damo221 Ilari: well, its an experimental kernel branch "alsa-kernel" im not sure if its the same as the main kernel22:15
sudoer I already commited it22:15
Ilari sudoer: In last commit?22:16
sudoer yes22:16
cehteh sudoer: remove that commit (anything else there to preserve?)22:16
Ilari damo221: Well then also specify where one got that kernel.22:16
sudoer Ilari: yes22:16
damo221 Ilari: some people have had trouble compiling just by cloning the latest git.... i was wondering if its possible to check out the identical revision which I used to write the module22:16
cehteh amend will do then22:16
Ilari sudoer: 'git rm --cached <bigfile>' and then 'git commit --amend'?22:16
cehteh and then gc/repack with some faith :) .. prolly expiring your reflogs first22:17
← Haarg left22:17
Ilari damo221: Well, there's 'git checkout -b <newbranchname> <id>'.22:17
damo221 Ilari: oooh, i can specify the ID of the actual revision and check it out as a separate branch?22:17
← suman left22:17
damo221 thanks22:18
Ilari damo221: And the checked out tree is the same one that you have...22:19
damo221 ilari: do i have to do a git clone of the repo first?22:20
Ilari: so git clone .... then git checkout -b newbranch <id>22:20
→ warthog9 joined22:20
→ merlin1983_ joined22:21
Ilari damo221: Yes. But if its some kernel fork, one could get it faster if one has existing kernel tree by using --reference or by adding new remote.22:21
damo221 Ilari: do i need to switch to the new branch after checkout?22:22
→ ashc__ joined22:23
← merlin83 left22:24
cehteh mhmpf why are submodules so pita22:24
damo221 nope apparently it switches automatically.... how do i switch back to the old branch22:25
cehteh git checkout22:25
← priidu left22:25
mugwump cehteh: sorry... it's been on the todo list to make them not suck for over a year22:26
→ bambule joined22:26
charon damo221: note that newer gits have 'git checkout -' as a short way to say "go back to the last branch"22:27
mugwump schacon: you have until 2009-11-09 until https://datatracker.ietf.org/idst/upload.cgi opens again22:28
cehteh you can not add recursive submodules to the toplevel you must add them truely recursive right?22:28
mugwump I don't really parse that question cehteh22:28
damo221 charon: thanks!22:28
cehteh and there is a long outstanding bug which barfs on adding a submodule with a relative path './'22:28
→ bdimcheff joined22:29
← thiago_home left22:29
cehteh i have mydir as master then submodule mydir/sub and mydir/sub/subnested22:29
i can not add this both submodules in order to mydir subnested must be a submodule for mydir/sub22:30
ok i guess i can live with that22:30
→ moonin joined22:30
← _patton left22:30
^aristotle charon: when was that added?22:30
that's awesome22:31
→ mattayers joined22:31
cehteh but always hand-editing .gitmodules to fix the paths after an add is bit unexpected22:31
moonin Does anyone know how to display a diff between staged versions in the index *without* including the working tree copy (and its conflict markers) at all?22:31
← mattayers left22:31
mugwump cehteh: yes, that's one of the bugs I've got on this list22:31
also switching branches doesn't change the submodule,22:31
cloning has no easy option to clone submodules too,22:32
cehteh that doesnt really hurt me yet22:32
Ilari The biggest reason submodules suck is that not having the right version checked out causes rebase and merge to abort.22:32
mugwump you also can't specify in .gitmodules that you prefer 'master' rather than the commit which was committed22:32
cehteh well i want only one repo and the submodules are branches in this repo, technically thats possible but no procelain for that22:33
mugwump yes, that's a valid use case I think22:33
→ warthog19 joined22:33
mugwump it's currently quite ugly22:33
← warthog9 left22:33
mugwump but really just a matter of cooking up the patches etc22:33
cehteh yes tracking refs than sha1's is high on the list for a lot people22:33
← ashc left22:33
charon 696acf45, first in 1.6.2-rc022:34
Gitbot [git 696acf45]: http://tinyurl.com/ykrfqrg -- checkout: implement "-" abbreviation, add docs and tests22:34
charon ^aristotle: ^22:34
cehteh the good thing is that it all can work, the bad thing is that the frontend sux22:34
mugwump that's right22:34
patches welcome, in this case help gladly given22:34
cehteh imo that would be a perfect gsoc project ... but summer is over :/22:35
← Dashkal left22:35
cehteh i am busy with other stuff .. but i had toughts about writing my own submodules porcelain22:35
khmarbaise Hi. I have questions. If i do a clone of a repository and create a branch and do a push. Does that branch continue to exist in the target repository?22:35
warthog19 → warthog922:36
→ jeckel joined22:36
Ilari Then wilder stuff I have plans for (but not all details on behavour are thought to end): empty branches (in way such that older clients won't break). Would simplify some stuff.22:36
cehteh khmarbaise: continue?22:36
moonin khmarbaise: When you say "target", do you mean the repository you cloned from and pushed to? If you pushed the branch, then yes it does.22:36
→ chovy_ joined22:36
^aristotle charon: :)22:36
chovy_ how do i revert a working copy?22:36
Ilari chovy_: 'git checkout -- <file>'?22:37
cehteh Ilari: well that are corner cases .. submodules are so much incomplete and buggy it slaps right into ones head sometimes22:37
moonin khmarbaise: I know that the default won't create branches on "git push" unless you explicitly list it onthe command line, but there are a number of ways to override it.22:37
← arrenlex left22:37
khmarbaise moonin: Yes "target" is the repository i cloned before from and pushed back to...So i have to explicitly push the branch?22:37
← tvw left22:37
moonin chovy_: "git checkout [<revision to revert to>] path/to/file22:37
chovy_ There are a couple of slightly different things you might mean by "revert" that I'd have to ask you about for more detail.22:38
chovy22:38
chovy_ it's just a local coyp, i haven't committed yet.22:38
copy*22:38
cehteh anything below a submodule is *always* ignored or can i force commit something on master what lies in a submodule22:38
chovy_ i modified ./foo and want to "revert" in the svn sense of the word.22:38
moonin chovy_: do you mean "to th last checked in version" or "to three versions ago" and do you want to revert *everything* or just a couple of files you screwed up?22:38
cehteh evil plan .. symlink submodule/.git to ../.git :)22:39
Ilari chovy_: 'git checkout HEAD -- foo'?22:39
cehteh tries :P22:39
chovy_ don't know. i'm a git n00b22:39
Ilari chovy_: (reverts it to last locally committed state).22:39
chovy_: HEAD => tip of current branch.22:39
chovy_ svn revert ./foo would wipe my local modifications to it, that's all i want to do.22:39
moonin chovy_: "git checkout foo" will do it. "git checkout -- foo" if "foo" might be mistaken for a branch name. HEAD is the default if not specified.22:40
Ilari moonin: It isn't default.22:40
cehteh git add -f chroot/.git22:40
fatal: Path 'chroot/.git' is in submodule 'chroot'22:40
Ilari checkout <file> defaults to staged version.22:40
cehteh bah it doesnt like me22:40
Ilari cehteh: And nothing in directory named '.git' can be added anyway. If you manipulate commits to contain such paths, checkout will fail.22:41
cehteh ok is there some other way to delegate a .git22:41
moonin Ilari: Uh... RTFMing to see if that's right...22:42
cehteh not reads but also writes/commits shall go up a alternatives/references22:42
→ s0ber joined22:42
moonin Ilari: Wow, you're right. My memory was wrong. It ends up being synonymous is a lot of cases, but thanks!22:42
← TeckniX left22:43
← [M7] left22:44
← alezandro left22:44
moonin khmarbaise: I don't often create upstream branches, so I may have this slightly wrong, but generally as a safety measure you have to specify "mybranch:[remotebranch]" to create a branch upstream. Once you have, though, it continues to exist until explicitly deleted even if you delete yours locally.22:46
→ Sigma joined22:46
khmarbaise moonin: Ah...thanks..22:46
← hagabaka left22:47
← Th0ma5 left22:47
→ newsham joined22:47
← agile left22:47
newsham I'm a little confused. If I "git checkout -b <somebranch>" does it create a new branch in my tree named that or does it pull a branch with that name form a remote archive?22:47
moonin khmarbaise: Of course, the --all and --mirror options also cause remote branches to be created, and if you list the branch in the config file, that's as good as on the command line... there's rather a lot of flexibility in the area.22:48
drizzd_ spearce: I have added smart HTTP support to t555022:48
moonin newsham: The former. It creates the branch (with the same contents as the current HEAD) and switches to it.22:48
drizzd_ spearce: would you like me to add that on top of my existing patches or should I squash it into the patch for t5540?22:48
newsham ok, thats not what I want to do.. how do I specify a branch to checkout/pull/whatever ?22:49
khmarbaise An other question. I have the following situaion: A Production line, for a prodcut already released (1.0) to a customer, may be need bug fixing on that and having a new release integration line (1.1), and a development line (2.0)...so the question is how to handle this in git ? Which kind of setup is needed ? Multiple Repositories ?22:49
moonin newsham: It's exactly the same as "git branch <somebranch> && git checkout <somebranch>"22:49
← makkalot left22:49
→ Azeroth-Work joined22:49
← tk left22:49
moonin newsham: The number of different things that "git commit" can do is somewhat confusing.22:50
spearce drizzd_: be curious to see what you have, but i just finished my own set of tests and was getting ready to send the respun series22:50
moonin khmarbaise: No, just multiple branches.22:50
newsham i am not really using git for changes/committing, I just want to reproduce certain src trees from a repo22:50
that coincide with different releases22:50
khmarbaise thx.22:51
moonin khmarbaise: Generally, you tag releases, and have branches for versions being worked on. You might have 2.0 development be "master" and 1.1 be "released" or "stable" or some such.22:51
drizzd_ spearce: ok, I'll wait for that then22:51
cehteh newsham: git reset --hard release would be bit rude then but works22:51
where release is eiher origin/branch or tagname or such22:52
← lohrun left22:52
newsham and that would fetch the particular release tagged as 'release' ?22:52
cehteh thats destructive to your master which becomes the intended version then22:52
→ lightcap joined22:52
→ richcollins joined22:52
drizzd_ spearce: it's funny how we now find bugs in the test suite because smart HTTP actually does some things right that dumb HTTP didn't :-)22:52
cehteh you cloned it .. you have to know what you want22:53
newsham so the clone url is the only thing that determines which versions of files?22:53
← damo221 left22:53
spearce drizzd_: :-)22:53
→ __iron joined22:53
cehteh newsham: clone gets you a copy of the repository22:54
moonin newsham: Git checkout <somebranch> will switch to and check out that branch. "git checkout <somebranch> <path>" will *not* switch branches, but will populate the <path> (recursively, so "." is quite powerful) with the given branch's files.22:54
drizzd_ spearce: does smart HTTP need USE_CURL_MULTI?22:54
spearce no22:54
← Pewpewarrows left22:54
← Turias left22:54
spearce actually, i had to bypass USE_CURL_MULTI, there was a massive performance bug caused by that code path22:54
→ Pewpewarrows joined22:54
drizzd_ that's what I thought, thanks22:54
spearce took me the better part of yesterday afternoon and today to debug that problem and find a workaround for when USE_CURL_MULTI was true at build time22:55
drizzd_ yeah, I never quite got my head around that one22:55
newsham ok, I did git clone of some repo.. then I did "git branch tag" and I *thought* I was getting sources from that branch that I pulled down22:55
now I want to play with a diff set of sources and I *thought* "git branch othertag" was switching to those versions of the files22:55
moonin newsham: To just check out particular versions, just us "git checkout <commit-ish>". A "commit-ish" is anything thta identifies a commit, like a SHA hash or a tag name or "tag^^^"22:55
newsham what do I need to do to undo the damage and get particular sets of flies22:55
→ tk joined22:55
drizzd_ Well, I'm glad you had time to do it. It's obviously not as easy as the spec made it sound...22:56
← ludde left22:56
newsham ahh, I was using "-b" I guess is the wrong thing22:56
moonin newsham: No, "git branch" just creates/deletes/modifies branch names. It never modifies the working tree files in any way.22:56
Ilari spearce: That 12M vs 300k bug?22:56
newsham oops, I misspoke.. I was using "git checkout -b <something>"22:56
spearce Ilari: yup22:56
→ Myrth joined22:56
newsham so I should have been using "git checkout <something>" ?22:56
moonin newsham: Ah, no don't do that. For just looking at sources, you should basically never use -b. Exactly.22:57
→ ereslibre joined22:57
← cannonball left22:57
newsham ok, how do I discard my stupid "-b" thing?22:57
moonin newsham: "git checkout <where you wanted to be>" and "git branch -d <bogus created branch>".22:58
Myrth hi, is there a quick way to convert a full copy to bare repo?22:58
← godsflaw left22:58
→ scarabx joined22:58
moonin newsham: Use "-D" if you get complaints from -d; it's the ""do it, damn it" version.22:58
Ilari Myrth: mv the .git out, nuke the rest and set core.bare to true.22:58
newsham ok, that seems to have done the trick.22:58
is there a way to query the repo for a list of branch tags?22:59
moonin Myrth: "rm -rf *; mv .git/* .; git config <something magic>"22:59
newsham (I can see some in git web browser page)22:59
Ilari newsham: ls-remote22:59
moonin Myrth: That's "git config core.bare true"22:59
jeckel → revans22:59
Myrth thanks!23:00
newsham ilari: nice. thank you23:00
moonin Myrth: Oh, and traditionally bare repositories are located in directories named *.git.23:00
→ qhoxie_ joined23:00
← jaysern left23:00
Myrth yep23:00
i love git23:00
moonin Myrth: That "rm -rf *" should be expanded to include any unwanted . files like .gitignore, of course; it was just shorthand for "blow away the working copy".23:01
← altrux left23:01
Myrth yeah i just moved out .git out of the folder and removed the folder23:01
newsham ok, I see one listed as "refs/heads/android-msm-2.6.29-donut" but if I "git checkout refs/heads/android-msm-2.6.29-donut" or "git checkout android-msm-2.6.29-donut" it says "pathspec did not match... did you forget to git add"23:02
what did I do wrong? :)23:02
← JKac3BEq left23:02
moonin Myrth: that will work; the "core.bare = true" thing just changes a few error checks to complain.23:02
Ilari newsham: They get fetched as remote tracking branches (git branch -r).23:02
spearce drizzd_: git://repo.or.cz/git/fastimport.git smart-http23:03
← struberg left23:03
← dunolie_ left23:03
moonin newsham: Remove the "refs/": git checkout android-msm-2.6.29-donut.23:03
← qrush left23:03
drizzd_ spearce: got it23:03
→ godsflaw joined23:03
Ilari newsham: Or 'git checkout -t origin/android-msm-2.6.29-donut'.23:04
Myrth is gitosis recommended, or there are better options?23:04
Ilari Myrth: gitolite23:05
newsham moonin: I did try that one (see above) and it failed23:05
ilari: "-t"? "--track"?23:05
moonin If I could ask my own question: what is the incantation to get "git diff" to show differences between staged unmerged versions, without showing the working tree copy at all?23:05
newsham unknown switch '-t'23:05
Myrth Ilari: thanks23:06
newsham ilari: with "git brach -r" do you mean "git branch -r <branch>; git checkout <branch>" or something like that?23:06
moonin newsham: If a file named ".git/refs/heads/android-msm-2.6.29-donut" exists (and is 41 bytes long and contains a SHA-1), it should work. Are you sure you don't have a typo?23:06
newsham moonin: I cut&paste to make sure no typos23:07
← __iron left23:07
newsham git ls-remote <url> |grep donut, cut&paste23:07
→ ctusar joined23:08
moonin Myrth: Note that you don't need anything if you have ssh set up, or git-daemon if you want to allow anonymous pulls.23:08
newsham: Oh! So that file doesn't exist! It's only on the upstream server! You have to fetch it.23:08
Myrth moonin: mmm... then what is the incentive of gitosis and gitolite?23:08
newsham how do I do that?23:08
charon moonin: git diff :2:foo :3:foo23:09
moonin: see the git-rev-parse manpage for what the numbers mean23:09
moonin newsham: If you cloned, it's already been fetched, but moved to the "origin/" sibdirectory. Try checkiung out "origin/android-msm-2.6.29-donut" before we get clever...23:09
(AFK for a few minutes...)23:10
doener moonin: (search for "stage number" in that man page)23:10
newsham moonin: origin/* seems to have done the trick (at least its taking a lot longer than the errors did :)23:10
→ bremner_ joined23:10
newsham ty for the help. this is very confusing.23:11
(believe it or not I have read through the git community book a month or two ago... *sigh*)23:11
→ alvarezp joined23:11
Ilari newsham: Local branches of another repository are mirroed as remote tracking branches.23:11
*mirrored23:11
→ d-snp joined23:12
Ilari newsham: Local branches are what you can check out.23:12
dwmw2 → dwmw2_gone23:12
newsham ok, after I "git checkout origin/android-msm-2.6.29-donut" if I run "git branch" it says I am now on "* (no branch)"23:12
is that normal?23:12
charon faq detached23:13
Gitbot charon: You are on a detached HEAD and might lose commits.. See http://git.or.cz/gitwiki/GitFaq#detached23:13
mugwump newsham: you want just 'git checkout android-msm-2.6.29-donut23:13
newsham it does look like I have the right code23:13
mugwump: doesnt work.. I get an error23:13
mugwump assuming you're using git from next23:13
charon newsham: and run 'git checkout -b android-msm-2.6.29-donut origin/android-msm-2.6.29-donut' to get a local branch if you want to do work on it23:13
mugwump otherwise try git checkout -t origin/android...23:14
newsham I dont plan on doing local work right now23:14
charon mugwump: isn't that an unfair assumption :)23:14
mugwump charon: who runs ancient released versions?23:14
newsham I have no '-t' flag. (git version 1.5.4.3 under ubuntu)23:14
← qhoxie_ left23:15
mugwump ok. git checkout -b android-msm-2.6.29-donut --track origin/android-msm-2.6.29-donut23:15
gitster unreasonable assumption, especially when you know you are helping somebody who gets confused by being on detached HEAD.23:15
charon mugwump: true, but then who hasn't known the ins and outs of git since early 2005 or so :p23:15
mugwump heh... successful troll :-P23:16
charon on the plus side, we finally have a rare case of confusion about fetch and remotes with pull not involved at all :p23:16
doener *runs*23:16
mugwump newsham: you only really need that if you are planning on committing to the branch23:17
→ VVSiz_ joined23:17
newsham mugwump: not planning to.. but good to know.23:17
→ mrlemao joined23:17
charon newsham: you should also upgrade your git if at all possible23:17
gitster If newsham does not plan to work on the code (other than reading and building), it is perfectly fine to do "git checkout origin/android..." and say "make" or whatever the build procedure is.23:17
newsham so, are gits sharp edges an intentionally barrier to entry to prevent a flood of kernel questions? ;-)23:18
mugwump otherwise, yes the detached HEAD message is right23:18
← nighthwk1 left23:18
mugwump maybe that could be made less scary ... hmm23:18
gitster exactly what part of this experience are you referring to as "sharp edge"?23:18
newsham anyway, many thanks for getting me through this.. I need to go back and read docs about this stuff some more23:18
← mrlemao left23:19
mugwump newsham: no, they're there to get people to come ask questions because we like talking to you23:19
drizzd_ spearce: you need to set LIB_HTTPD_PORT before you run lib-httpd.sh (otherwise it will set 8111 internally)23:19
newsham gitster: the process of trying to get a particular tagged version of the remote source tree23:19
spearce drizzd_: thanks23:20
drizzd_ spearce: but I guess I will comment on it when you send it to the list23:20
charon newsham: tags are global of sorts though. origin/* is not a tag, but a representation of the _branches_ of the 'origin' repository.23:20
gitster newsham: meaning that the message you get when you run "checkout origin/android..." is unnecessary?23:21
← VVSiz left23:21
← sgh left23:21
newsham gitster: I had no idea about how to use "git checkout origin/..." in the first place23:21
moonin newsham: It's quite straightforward; anything that looks vaguely like a revsion can be checked out. The problem was that you were looking at the upstream server's list of branches, and git-clone adds "origin/" in front of those to avoid cluttering up your private list of branches.23:21
← onigiri left23:21
newsham "its quite straightforward [...] the problem was [...]" :)23:21
moonin The confusion was really about what "git clone" does rather than "git checkout".23:21
newsham that clarifies.. thank you.23:22
gitster newsham: that's a bit unfair use of the word "sharp edge", I guess.23:22
newsham I wouldnt go as far as to say "its straightforward". but it is definitely seeming more logical now.23:22
gitster: I was trying to be a little tongue in cheek.. my apologies.23:22
mugwump trolling us eh ... what's the commands for p4 again?23:23
newsham but I have to be candid -- this does seem a bit more complex than many revision control systems I've used23:23
moonin newsham: The "-t" flag turns on some additional features for people *developing* on that branch. If you're not planning on adding any commits, -t is entirely superfluous.23:23
gitster no need to apologize. I am lurking to find true sharp edges, and occasional false positives are expected.23:23
→ niki joined23:23
newsham moonin: my cmd line version doesnt seem to support -t. Am I on an ancient rev?23:23
mugwump that should be --track23:23
doener moonin: additional features? It simply causes git to create a branch head23:23
gitster slightly older. What does "git version" say?23:23
doener moonin: at least in the "git checkout -t origin/foo" form23:24
newsham 1.5.4.323:24
moonin doener: Oh, doesn't it also record the trasking relationship in .git/config?23:24
charon on a completely unrelated note, hg-git worked very well for me so it seems we now have a free choice of hg/git frontend for hg/git repos23:24
gitster moderately older, then ;-)23:24
mugwump that's like 22 months old :)23:24
doener moonin: yeah, but that's the default if the starting point is a remote tracking branch anyway23:24
mugwump newsham: newer versions are worth it if only for the better docs23:25
doener moonin: i.e. "git checkout -b foo origin/foo" also does that (at least by default)23:25
gitster doener: if your git is new enough ;-)23:25
← Diapo left23:25
newsham thank you, I'll pull down a newer version.23:25
mugwump prefers 'git checkout foo' for that ... yay it got in!!23:25
Myrth how to apply a .patch file on git copy?23:25
← warthog9 left23:25
Ilari Fortunately release of Lenny drastically reduced the frequency one would run into 1.4.4.4...23:25
Myrth i see that it's possible to apply from mail23:26
git am23:26
newsham (this is what I got from apt-get on ubuntu 8.04.2, btw)23:26
mugwump Myrth, you can use git apply23:26
← rudi_s_ left23:26
mugwump newsham: yes, ubuntu stable23:26
Myrth that was simple :P23:26
doener gitster: yeah, but it's been a while since I saw someone with a git version older than 1.5.323:27
moonin Myrth: Or just "git apply". "am" is somewhat more automated and fills in a commit comment and does the commit.23:27
mugwump I'm running ubuntu unstable (karmic) and it has 1.6.3.323:27
moonin mugwump: Karmic indeed! Can you e-mail gitster a copy of the source so he can avoid having to do any development work for the next year? :-)23:28
← dsch04 left23:28
← hobodave left23:28
mugwump missed the ojoke23:29
drizzd_ spearce: t5551 usually passes, but sometimes I get FAIL 3 "-< Transfer-Encoding: chunked\n+< Content-Length 256"23:30
moonin Is there an easier way to get a diff between the staged versions of an unmerged index file than "for i in 1 2 3; do git show :$i:path/to/file.c > /tmp/$i; done; diff -u /tmp/[12]...23:30
← Theravadan left23:30
spearce drizzd_: whoa, both a transfer-encoding chunked and a content-length header from apache?23:30
← bremner_ left23:30
charon [00:08] <charon> moonin: git diff :2:foo :3:foo23:30
→ vinse joined23:30
drizzd_ spearce: no, you're expecting chunked transfer-encoding, but you're not getting it23:31
moonin I'm really f*ing sick of conflict markers showing up in my diffs when I just want to see "what changed on side 1" and "what changed on side 2" of a merge...23:31
spearce oh, probably because its so small of a response. ok.23:31
mugwump moonin: see man git-mergetool23:31
Gitbot moonin: the git-mergetool manpage can be found at http://git.or.cz/man/git-mergetool23:31
gitster is amazed by charon's example---how intuitive and straightforward for somebody who knew about "git show :$i:path/to/file" it is!23:31
moonin (I have a messy one with *lots* of conflict markers, so it's hard to read.23:31
← tbuehlmann left23:32
charon moonin: note that you can also 'git checkout --conflict=diff3 -- $file' to see the base version in the conflict hunks23:32
jonthefly so ive read through the git-hook man page, Im trying to create a hook that will get executed every time a git pull is performed.23:32
charon gitster: i was rather amazed that it worked, too ;)23:32
jonthefly there is post-merge, but this hook isnt executed if git-merge isnt called23:32
→ Theravadan joined23:33
gitster Please justify why 'echo "git pull ; .git/hooks/post-pull" >my-git-pull' is insufficient.23:33
→ cpg joined23:34
gitster By the way, today's 'pu' does not pass self-tests; pu~2 does, though.23:34
moonin charon: Now, how do I extend that to a git-style 3-way diff....23:34
After trying "git mergetool", I have a new problem: How do I get back the staged files after it "git add"ed the version with the conflict markers....23:35
cpg hi! .... i am in a branch that is tracking a remote branch, and i am not sure how to pull things in ... here is the (unusually verbose) message i get from git ... http://www.fpaste.org/ujrY/ .... i want to bring in the latest in the head of the (remote) fedora-10 branch23:35
jonthefly id idnt know there was a post-pull :)23:35
oh are you saying i should create a bash script?23:36
cpg but i don't really know exactly how ... (yes, i did read some tutorials, etc, but this part on handling remote branches i did not understand well)23:36
jonthefly I have users who are too retarded to remember a bash script23:36
to remember to use it....23:36
→ Dashkal joined23:36
jonthefly I want to ensure that when they git pull that a certain command gets executed, im actually thinking that maybe git-submodules is what i need to look into23:36
← newsham left23:37
charon moonin: well, you can apparently do 'git show :1:foo > foo; git diff-files --cc'23:38
don't think you can restore the unmerged index without redoing the merge, though23:38
Ilari cpg: 'git pull origin fedora10'23:38
cpg: (or whatever the branchname was).23:39
→ hobodave joined23:39
cpg Ilari: oh, thanks!!23:40
← nothingHappens left23:40
charon gitster: sadly 'git diff :1: :2:' doesn't work, as it doesn't refer to some virtual tree of all the stage-1 entries23:40
gitster The example was about :1:foo :2:foo wasn't it?23:41
charon yes, that was more a tangent along the lines of surprisingly intuitive features23:41
gitster My earlier comment was ":1:foo works in one place and naturally it works somewhere else". ";1:" itself consistently does not work anywhere ;-)23:41
charon true :)23:42
gitster s/work/mean anything/;23:42
qrush|away → qrush23:42
moonin charon: That would be really nice. I compressed it with "git diff :{1,2}:path/to/conflicted/file.c", but what I was really hopig for was a23:44
Myrth: "rm -rf *; mv .git/* .; git config <something magic>hopig for was a "git diff -c" style diff of all *three* versions. I find that a *lot* more legible than "diff3".23:45
→ Etherael joined23:46
Etherael is there a way to search the entire history of a git project for the existence of a specific file?23:46
charon moonin: as i said, you can get that by loading the stage 1 entries into the worktree first and using 'git diff-files --cc'23:47
moonin Etherael: Yup. "git log -- filename" will show every commit that modifies that file name, including deleting it.23:47
Etherael super, thanks23:47
charon moonin: there's 'git ls-files -u' to find the unmerged entries, if you want to script around it.23:48
moonin charon: *smacks head* D'oh! Yes, that's what I should have done. Move the conflict marker version aside, check out the :1: version, and diff-files will do what I want.23:48
Kind of a kludge, but it'lll do.23:48
charon you don't need to move away the conflict marker version, 'git checkout --merge -- $file' will restore it.23:48
← ia left23:48
Etherael Hmmm.23:48
dbt is there a way to change the gitk presentation?23:48
Etherael wonder if that messes up msysgit, it dies whenever I do git log -- (anything goes here)23:49
moonin charon: Ah, thanks for the tip. I thought there was a way to do that, but couldn't think of it off-hand.23:49
cehteh mugwump: git submodule init should know --recursive too23:49
Etherael just silently does nothing23:49
charon Etherael: that means no commits modified the paths you specified23:49
moonin Etherael: Um, if there's nothing to list, it can spend a while searching history and producing nothing. Kind of like "grep -r string_not_found". I suspect it's working.23:49
charon in general, an empty 'git log <arguments>' means none of the commits in the current branch match the <arguments>23:49
Etherael charon: it just instantly fails23:50
well23:50
→ ablemike joined23:50
Etherael maybe not fails23:50
but if it's finding nothing, it's finding nothing awfully fast.23:50
moonin Etherael: How big is the project?23:50
Etherael moonin: 700mb23:50
charon Etherael: git is so blazingly fast that it indeed is frequently instantly, and silently, successful23:50
← patrikf left23:50
Ilari Etherael: 'git log -- <file>'?23:51
Etherael I can only make it do something with git log -- * which appears to be the same output as just git log23:51
moonin Etherael: I mean, how many commits? If it's less than a couplf of thousand, git will do it in less than a second. It takes 1.227 seconds elapsed for me to search the Linux tree.23:51
Etherael 1115 commits23:52
(well, revisions, that's the same thing isn't it?)23:52
← Ryback_ left23:52
→ futoase joined23:52
Etherael but say for example I'm looking at a file in the root dir of the git repo that I can see is under version control23:53
moonin Etherael: Note that the "*" there probably doesn't do what you want. It's expanded to the current directory contents.23:53
← ablemike left23:53
→ marze joined23:53
← ereslibre left23:53
→ hugoxrosa joined23:53
Etherael ahh, ok, actually now it does appear to do stuff.23:53
Ilari Etherael: Execute commands in root of the repo. What commands run in subdirectory do depends on phase of the moon...23:53
Etherael all I have here is a filename, I don't know where it should actually sit within the directory hierarchy, is there a way to do real wildcards without them being expanded out by the shell?23:54
cpg hi all after pulling properly, i cannot push now ... http://paste.amahi.org/d636de4fb however, i am not sure why, as i knly did a cherry pick of a committ that was quite clean (a one liner) and i see no reason why ... -v does not help :(23:55
→ dunolie joined23:55
cpg any idea Ilari ?23:55
Ilari: the repo is at http://git.amahi.org/?p=amahi.git;a=summary23:56
moonin Etherael: Ah, that's a bit tricker. I'm RTFMing to try to find an answer...23:57
Ilari cpg: 'git checkout fedora-10', 'git pull origin fedora-10', 'git push origin fedora-10'?23:57
Etherael moonin: Thanks! I'll race you ;)23:57
← futoase left23:58
charon Etherael: not sure it's even possible. i would try 'git log --oneline --name-status | less +/filename' if i had this problem.23:58
cpg hi Ilari - still fails ... i wonder if it's because we have a branch called fedora-10 and a label at the base of the branch also called fedora-1023:59
i had to do git co origin/fedora10 (otherwise the tag (not label) was picked up)23:59

Logs Search ←Prev date Next date→ Channels Documentation