IRCloggy #git 2008-11-22

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.

2008-11-22

← david_koontz left00:00
askentask anyone?00:02
apeiros_ felipec: even if I don't like that answer, thanks :)00:03
doener askentask: what about the other question you got? Did you upload your ssh key to github? And is ssh actually using it? (Either because it has a default name, or because you configured ssh to use it)00:04
← ankit9 left00:04
felipec apeiros_: check the .gitmodules format... it specifies a url... nothing else00:06
apeiros_: so I guess you can use whatever git clone can use00:06
askentask: ok, your url looks fine, now you need to make sure your ssh public key is configured in github00:07
→ peeja joined00:10
← hobodave left00:11
← jackdempsey left00:11
← gottesmm left00:12
← SRabbelier left00:13
← Tv left00:17
→ carllerche joined00:18
askentask ssh?00:18
secure hell right?00:19
no i havent, how do i make one?00:19
in git or another program?00:19
← brett_h left00:21
felipec askentask: I'm sure it's in the github faq00:21
partha → partha-bbl00:22
← _graham_ left00:22
felipec askentask: http://github.com/guides/providing-your-ssh-key00:22
→ akitada joined00:25
← lfaraone left00:27
→ willb joined00:27
← exon left00:27
← kumbayo left00:29
askentask thanks00:31
which is the exact key: shouldProc-Type: 4,ENCRYPTED00:31
DEK-Info: DES-EDE3-CBC,E4655FA109221D3F00:31
be in?00:31
or just the:00:31
41keBtyx/sOYzQPreg4BgUK/bDZIAcPiWvMzdzq/wYyImMDh5DUVetbS2bBJi7kR00:31
rXAn16Xf00:31
...00:31
part00:31
doener askentask: wrong file, that's your private key. Nobody but you is supposed to know that one00:32
offby1 wait, I'm still copying it down; wait, don't clear the screen00:32
askentask lol00:32
deskin gitlog ftw :)00:33
← apeiros_ left00:35
→ apeiros_ joined00:35
askentask no i get:00:36
ssh: no address associated with name00:36
← careo left00:38
← icwiener left00:38
← Bass10 left00:38
← sdboyer left00:39
← drewr left00:40
← nud left00:41
askentask noone?00:41
→ apeiros__ joined00:41
← apeiros_ left00:42
cbreak askentask: well... what does ping do?00:42
apeiros__ → apeiros_00:42
← Jacolyte left00:42
cbreak it looks like the name you picked can't be resolved00:43
bremner or host, since ping can hang for other reasons.00:43
probably just a typo00:43
cbreak well, if ping can't resolve the name, it won't hang :)00:43
askentask now i at least got to type my pw but then the remote end hang up again00:46
← CodeOfficer left00:49
askentask error: src refspec master does nto match any00:50
error: failed to push some refs to00:50
doener sounds like you don't have a branch called "master"00:51
or maybe you have an ancient git version. Does "git push origin refs/heads/master:refs/heads/master" work?00:51
→ gottesmm joined00:55
← ciskje left00:55
← Sadr left00:55
askentask wtf diff errors each time00:56
→ CodeOfficer joined00:56
← a2800276_ left00:57
doener askentask: I guess you only did "git init" but did not commit yet?00:57
askentask doener: that fails too00:58
should my email eb at the end of the key or should i delete that?00:59
← pcapriotti left00:59
doener that shouldn't matter IIRC00:59
askentask: what fails too?00:59
← stephens left00:59
askentask git commit -m "first commit"01:00
doener did you "git add" anything?01:00
← loincloth left01:00
askentask git push origin refs/heads/master:refs/heads/master fails too01:01
src refspec does not match any01:01
doener if you didn't commit yet, that's expected, there's simply no branch yet01:01
← prof_freud left01:03
→ Tv joined01:05
← felipec left01:05
→ Jacolyte joined01:06
askentask what exactly do you commit with?01:09
i added adn committed once as int he sintructions01:09
then i added shapegen.clj and now i want to commit it, how?01:09
→ softdrink joined01:09
doener git add shapegen.clj; git commit01:09
askentask what does -m mean?01:09
jql commit [m]essage01:10
doener You can give a commit message on the command line with that01:10
but writing a useful commit message with -m is rather painful01:10
without -m, you get your editor started to write the commit message01:10
jql I commit often enough that I still leave a chapter's worth of commit messages per day01:10
even with -m01:11
askentask how can i check what is added and what is committed?01:11
doener git status01:11
git diff01:11
git diff --cached01:11
that's status (what a surprise!), a diff of the unstaged changes and a diff of the staged changes01:12
← peeja left01:13
bobesponja → bobesponja_comie01:15
bobesponja_comie → bobesponja_comer01:15
→ manunderground joined01:17
← pd left01:18
← sgarza left01:20
askentask nothing added to commit but untracked files present01:20
← ebzzry left01:21
askentask when i add file to /progs/gitinit-stuff/01:21
i am in progs/ and do git add shapegen.clj01:21
then cd into git-init-stuff?01:21
→ lordpil joined01:23
→ ebzzry joined01:23
askentask fakk this shit01:24
← askentask left01:24
← Yuuhi left01:25
→ wrobbie joined01:28
← cgardner left01:29
→ sgarza joined01:29
→ pd joined01:31
← sgarza left01:32
← kennethlove left01:35
← CodeOfficer left01:35
← x_or left01:35
bobesponja_comer → bobesponja01:36
→ jrb joined01:37
jrb is it somehow possible to copy .git/config when cloning a remote repository?01:39
doener no01:39
TML It seems unlikely that would ever be the right thing to do...01:39
jrb yes a lot of information in there would obviously be wrong01:40
would have been interesting though01:42
← bronson left01:42
→ tclineks joined01:42
tclineks if i have staged changes, how do i make the working tree look just like what's been staged01:43
doener yeah, like having someone cloning a config with commti aliased to "rm -rf $HOME"01:43
tclineks realize i could reverse-patch git diff HEAD but seems ugly01:43
doener tclineks: git checkout -- .01:43
tclineks: that reverse patch would give you a working tree equal to HEAD, not equal to the index (as you have staged changes)01:44
tclineks ok01:44
thanks01:44
doener tclineks: with recent git versions, you can also do "git stash --keep-index"01:44
tclineks ah that's nice01:44
doener that stashes only your unstaged changes01:44
→ CodeOfficer joined01:49
tclineks how would i make workign tree look like HEAD but keep index the same, reset --soft?01:49
jast no, reset --soft leaves the working tree alone01:49
there is no straightforward way to "skip over" the index when restoring stuff01:49
doener tclineks: as I said, that's what the reversed patch would do01:49
jast that's because the commands that bulk-update the working tree get their information out of the index01:49
→ rwp joined01:49
tclineks hm01:49
ok01:49
jast i'd just save a temporary diff of the index somewhere, do a reset --hard and reapply the diff01:49
hmm01:49
→ netoman joined01:49
jast that doesn't really solve the problem, does it :)01:49
← kanru left01:50
→ kanru2 joined01:50
deskin you can specify a temporary index file, how does that go? GIT_INDEX_FILE=tempfile git read-tree HEAD; GIT_INDEX_FILE=tempfile git checkout --01:50
jast so here's a simply but ugly solution: mv .git/index{,2}; git reset --hard; mv .git/index{2,}01:51
doener if you don't use the autocrlf stuff, you could also do: git diff -z --name-only HEAD | xargs -0 -i bash -c 'git show HEAD:{} > {}'01:51
deskin's version is probably the best one01:51
jast yeah01:51
though i suppose you could make it a bit shorter01:51
doener yep, GIT_INDEX_FILE=tempfile git read-tree -u --reset HEAD01:52
jast how about: GIT_INDEX_FILE=foo git reset --hard01:52
deskin please do :)01:52
doener jast wins01:52
jast yay01:52
→ rmh3093 joined01:52
jast now i'm rich and famous01:52
deskin always enjoy a good round of git golf01:52
jast i was going to suggest GIT_INDEX_FILE=/dev/null but i don't know if git would like that01:53
doener not unless you have write permissions for /dev01:53
jast and then bad things would happen01:53
← moccuo left01:53
doener right01:54
deskin not sure you can mmap /dev/null01:54
← priidu left01:54
doener deskin: git creates the index file from scratch as $GIT_INDEX_FILE.lock anyway01:54
deskin ah, I see that now too01:54
doener and then replaces the old $GIT_INDEX_FILE with the new one01:54
tclineks GIT_INDEX_FILE=foo git reset --hard is awesome01:55
deskin bad bad bad01:55
tclineks thanks jast01:55
← anttih left01:56
tclineks doener why `git checkout -- .` vs just `git checkout .` ?01:56
doener tclineks: -- disambiguates tree-ishes and paths. Although useless in that case, I've made it a habit to always use it01:57
tclineks ok that's waht i though01:57
t01:57
doener wonders when the DDoS against InternetX will finally end01:58
doener I'm sick of hunting ip addresses and adding them to my /etc/hosts...01:59
→ sgarza joined01:59
← aspotashev left01:59
← tvw left02:01
→ tvw joined02:01
→ slava joined02:01
slava how do i get git to not convert line endings?02:01
i'm on windows but i want \n to remain \n02:01
it seems that git-clone converts them02:01
doener slava: git config core.autocrlf false02:02
slava but they get converted when i clone02:02
do i set the config in the upstream repo?02:02
← rmh3093 left02:02
doener slava: doing a "git reset --hard" should fix things up after the git config command02:02
slava: you can also set it in the global config file, using --global02:03
slava but i don't want to do this every time and i have control over the upstream repo02:03
→ rmh3093 joined02:03
jast the upstream repo doesn't get to decide02:03
however, you can set this in your local user-specific git config so it's valid for all of your local repos02:03
slava will git config --global on the windows client do the trick thne?02:03
jast yes02:03
slava great02:03
thanks, this had me stumped because on another windows box it doens't do this, but i guess its msysgit -vs- cygwin git02:04
doener yep02:04
IIRC msys defaults to autocrlf on, and cygwin to off02:04
slava thanks for the help02:05
← slava left02:05
→ jesselucas joined02:05
rwp I would like to reduce the size of a repository by discarding a branch with large useless binary data. How?02:06
→ coderdad joined02:06
→ ilogger2 joined03:33
→ softdrink joined03:34
→ kanru joined03:44
→ manunderground joined03:51
→ roylee17 joined03:51
→ ben_h_ joined03:54
→ evalsaidem joined03:56
→ mikearr joined04:01
→ elmex_ joined04:02
← mikearr left04:02
→ dreiss joined04:05
→ mikearr joined04:10
→ careo joined04:11
→ muthu_ joined04:14
elmex_ → elmex04:17
← ben_h_ left04:26
← careo left04:30
→ quoin joined04:32
← quoin left04:36
→ quoin joined04:40
← drio_ left04:42
→ drio joined04:42
← manunderground left04:45
→ jesselucas joined04:45
→ fhobia joined04:53
→ ben_h_ joined04:53
→ scientes joined05:06
→ moccuo joined05:14
→ sdboyer joined05:15
← moccuo left05:15
→ a2800276 joined05:15
← dreiss left05:22
→ tcoppi joined05:22
→ perezd88 joined05:35
← perezd88 left05:41
→ coderdad joined05:41
→ skrit joined05:45
→ careo joined05:47
→ unreal joined05:48
→ ebzzry joined05:55
← ebzzry left06:00
→ loincloth joined06:00
→ agile joined06:03
→ Fla99 joined06:04
Fla99 how do I change the destination where my git push commands pushes to?06:06
jql git config remote.<name>.push <refspec>06:08
werdnum hello06:09
jql git-config(1) and git-push(1) give various options that might be worthwhile06:09
werdnum there's a branch newer than my git svn checkout which doesn't seem to exist in my local git copy - git svn find-rev r43712 gives nothing.06:10
Ilari Fla99: You mean change the upstream repository URL?06:16
Fla99 I found it. it's in the config file06:16
Ilari Fla99: The setting name would be 'remote.<name>.url'.06:17
Fla99 i meant the url attr under remote ["origin"]06:17
Ilari?06:17
is that something different from what I mean?06:17
Ilari Fla99: You mean '[remote "origin"]'? The url setting under it is named 'remote.origin.url'. git config takes those names...06:19
← careo left06:19
Fla99 ah06:19
you mean that is a setting you can set with a command06:19
with git config06:19
i just manually edited the file06:19
guess would be nicer with the git command06:19
Ilari Fla99: Try 'git config repository.origin.url'. It prints value of that setting.06:19
Fla99 k06:20
Ilari Fla99: Aargh... 'git config remote.origin.url'.06:20
Fla99 ah06:20
yeah works perfectly06:20
now you can add with git config remote.origin.url --add ssh://mynewrepos.com/app.git06:21
is that what u are saying?06:21
or with value06:22
→ GNUix joined06:22
Ilari Fla99: Don't use '--add' with settings like 'remote.foo.url'. It could be useful with settings like 'remote.foo.fetch' or 'remote.foo.push'.06:23
→ dreiss joined06:24
Fla99 k06:24
Ilari Fla99: If you use '--add', it doesn't overwrite old setting (creating one if it doesn't already exist), but forces creation of another setting with the same name.06:25
→ careo joined06:25
Fla99 yeah makes sense06:25
so value is the way to go?06:25
Ilari Fla99: It's 'git config remote.origin.url <URL>'. That properly overwrites the value...06:26
Fla99 ah06:26
ok06:26
thanks Ilari06:26
Ilari Fla99: --add is only to be used with settings that can appear multiple times (like 'fetch' and 'push' settings of remote).06:28
→ rubydiamond joined06:28
→ mediogre joined06:29
← unreal left06:32
→ unreal joined06:32
→ aspotashev joined06:33
← softdrink left06:33
→ peeja joined06:38
← jesselucas left06:40
← agile left06:44
→ mediogre` joined07:00
werdnum Hi! I have a folder in .git/info/excludes, I've marked all files in it with git update-index --assume-unchanged, and it STILL shows up as me having deleted that file every time I delete the folder and git diff07:01
Ilari werdnum: --assume-unchanged only makes sense for files that are tracked, and tracked files can never be ignored.07:02
werdnum then how do I tell it to ignore me deleting a file?07:03
Ilari werdnum: The file is supposed to exist in repository but not in working tree?07:03
werdnum yeah07:04
← mediogre` left07:04
← unreal left07:05
Loof I've got an odd issue... I'm trying to resolve a merge just using the changes from the branch... so I'm using checkout-index to get the stage 3 version07:06
but when I do an add, I get 'unable to stat (No such file or directory)07:06
the file is there07:06
so I'm rather confused07:06
→ unreal joined07:06
→ Flathead joined07:07
← mediogre left07:07
werdnum Ilari: were you about to answer, or don't know?07:10
← muthu_ left07:12
werdnum Looks like git update-index --remove did the trick07:14
Ilari werdnum: Except that update-index --remove killed the index entry for that file, so when you do next commit, that file won't be in it anymore...07:15
werdnum but git diff gives me no changes07:16
damn, git commit says not.07:16
← dreiss left07:16
Ilari werdnum: 'git status' can be used for dry-run of 'git commit'. It even accepts the same options (for options that control what is committed). Try stuff like 'git status -a'.07:17
werdnum So *is* there a way to do what I want to do?07:18
Ilari werdnum: I don't know (assuming unmodified version of git)...07:19
werdnum it's just a symlink, so there isn't that much stress.07:19
it's just.. annoying :)07:19
Ilari werdnum: Hmm... What does 'git diff' show for file if diffing is disabled for that file?07:20
werdnum diff --git a/phase3/extensions/README b/phase3/extensions/README07:21
deleted file mode 10064407:21
index 85f4943..000000007:21
etc07:21
Ilari werdnum: Ah, it's that 'binary files differ' (with diff --git and index lines)...07:22
werdnum no07:22
Ilari werdnum: 100644 means regular file.07:23
werdnum just --- a/phase3/extensions/README07:23
+++ /dev/null07:23
→ Weasel[DK] joined07:23
werdnum Ilari: yeah, it's a README..07:24
Ilari Loof: Strace it and look what is failing?07:24
Loof not yet, but I'll give up this for t'nite and try that in the morning07:25
Ilari werdnum: It seems that --asssume-unchanged does affect 'git status' but not 'git diff'.07:27
werdnum does affect git commit -a, but not git commit07:28
Ilari werdnum: 'git commit' commits the contents of index anyway.07:29
werdnum mmm07:29
Ilari werdnum: So it doesn't matter if the working tree file is deleted or contains whatever garbage. 'git commit' commits the version in the index.07:30
werdnum mmm07:31
now how do I make it ignore changes to the working tree (keep them out of index, even with -a)?07:31
Ilari werdnum: It seems that 'commit -a' also ignores assume-unchanged bit...07:31
werdnum yeah07:32
that's kinda annoying :)07:32
Ilari werdnum: It seems you can't. But that's what 'git commit' (without specifying anything to commit) is for.07:32
werdnum mmm07:32
ah well, thanks.07:32
Ilari werdnum: The usual workflow is to first stage whatever changes you want and then call 'git commit'. There are ways (such as 'git add --patch') to stage some file with contents that it has never had on-disk.07:34
→ scook0 joined07:35
werdnum mm07:35
Ilari werdnum: If you like graphical interfaces more, 'git gui' can also be used to perform partial adds of files...07:35
werdnum spiffy.07:36
← skrit left07:36
Ilari werdnum: Both can at least add changes chunk-by-chunk. IIRC, latter versions can even do line-by-line...07:36
werdnum ooh, that's awesome!07:37
← fhobia left07:38
← scientes left07:39
→ scientes joined07:40
← a2800276 left07:45
→ a2800276 joined07:46
→ scientes_ joined07:46
← scientes left07:46
← coderdad left07:49
→ skrit joined07:49
→ ashleyw joined07:52
→ coderdad joined07:53
scientes_ → scientes08:01
→ ezyang joined08:03
ezyang Hello all; when I capture the output of `git fetch`, I get a long list of each percentage that is dynamically output. How do I turn off this "status output"?08:04
(I'd still like informational/warning messages. Just not the interactive elements)08:04
.08:04
actually, on second glance, it doesn't look like this is Git's fault08:06
← rubydiamond left08:06
→ rubydiamond joined08:07
→ carllerche joined08:07
→ thiago_home joined08:08
← rubydiamond left08:10
→ rubydiamond joined08:10
→ daaku joined08:10
← careo left08:13
→ CelticSoul joined08:14
→ pheaver joined08:18
pheaver I'd like to search a repo's entire history for all files that08:20
have some property (for example, all files larger than some08:20
size). How can I do this?08:20
→ Ramblurr joined08:24
aspotashev ezyang: I don't have stats on "git fetch": http://pastebin.com/ddb790b208:27
ezyang: do you mean "git pull"?08:27
ezyang ezyang: Yah, that's what I mean. Not gits fault :-)08:27
→ stuffcorpse joined08:28
→ ToxicFrog joined08:29
aspotashev ezyang: then, use "git pull --no-stat"08:29
ezyang No, I mean, it's entirely not git's fault. A different command was giving the spurious data08:29
Ilari pheaver: You can use 'git rev-list --objects' to get list of objects involved in history. Then see what objects are blobs (those and trees always have path after their ID in rev-list output) and check if those have the specific property.08:30
pheaver hmm... ok08:30
← stuffcorpse left08:30
Ilari pheaver: Fast way to filter list of IDs for blobs is to use 'git cat-file --batch-check'.08:30
pheaver great, thanks!08:31
Ilari pheaver: Sadly, it doesn't like those paths in output, so you need to save them...08:31
pheaver oh08:31
Ilari pheaver: And if size is involved, --batch-check also gives the size of file...08:32
pheaver: It gives output like 'b2dc968e60c94189805228f5d99786bd50189583 commit 1036'. Meaning object 'b2dc968e60c94189805228f5d99786bd50189583' is of type 'commit' and is 1036 bytes in size. For objects of type 'blob', the blob size is equal to the file size (it is stored raw).08:33
Gitbot [git b2dc968e6]: http://tinyurl.com/5szgny -- t5516: refactor oddball tests08:33
pheaver so, if i do git rev-list --objects, then filter out only the commit ids which have a path listed after them08:34
→ kandjar` joined08:34
kandjar` hi there08:34
pheaver oh wait, that's what the cat-file is for08:35
kandjar` I m encountering an issue with am (which I think make sens, but i d like to get around it)08:35
basically, I have 2 repo, one with the changes from A to M and the other one with the changes from M to Z, I want to merge both repo08:35
what I did is: in the second repo: git format-patch master root -o patches08:36
this worked fine. then in the first repo I tries: git am ../second/patches08:37
but this gave the output: 'Nothing to do'08:37
pheaver kandjar`: I would think that you could just create a branch in one repo that tracks the other, then just merge or cherry pick08:37
Ilari pheaver: The object IDs that have paths are not commit IDs. They are tree or blob IDs.08:38
pheaver oh08:38
← carllerche left08:39
Ilari pheaver: There are also tag IDs, but they don't appear in rev-list output (since commits can't refer to tags).08:39
kandjar` pheaver: what do you mean?08:39
Ilari pheaver: Git has four types of objects 'commit', 'tree', 'blob' and 'tag'.08:40
pheaver kandjar`: well, from your A-to-M repo, you can create a remote branch that tracks the other repo08:41
kandjar` but it's assuming that they start at the same level no?08:41
pheaver ?08:41
Ilari pheaver: In short: Blobs store contents of files, trees build them into directories, commits bind them into history and tags introduce trust into system.08:41
kandjar` can I just to: git remote add m_to_z secondrepo?08:42
pheaver kandjar`: yup!08:42
kandjar` where will the second repo be linked?08:42
pheaver i don't know what you mean08:42
kandjar` I mean: how will both repo be linked together?08:42
they must have a common parent08:43
no?08:43
pheaver no, they don't08:43
they are two unrelated branches08:43
which is entirely possible08:43
kandjar` oh?08:43
pheaver yeah08:43
→ ajonat_ joined08:43
pheaver so then you just do something like git branch -r bla m_to_z/master08:43
and now you can merge bla into master, hopefully08:44
well ok, i guess you wouldn't be able to merge08:44
← scook0 left08:44
pheaver but at least cherry pick08:44
Ilari pheaver: Why the '-r' there?08:44
pheaver isn't that the right syntax for tracking a remote branch?08:45
Ilari pheaver: Depends on nature of the trees. Sometimes merge does work. Sometimes it produces horrible amounts of conflicts...08:45
pheaver yeah, but cherry pick should work fine in this case08:45
kandjar` pheaver: indead I can cherrypick08:45
Ilari pheaver: That's 'git branch --track bla m_to_z/master' or in later versions 'git branch bla m_to_z/master'.08:46
pheaver kandjar`: yay :)08:46
kandjar` but I have 70 changeslists08:46
anyway to automate the final stuff?08:46
pheaver well 'git branch -r bla m_to_z/master' is how i've been doing in for a few weeks now and it works for me :)08:46
kandjar` is there any way to ease the process?08:46
pheaver kandjar`: yeah, i'm sure you can script this08:46
git cherry will yield a list of commits that are in one branch but not the other08:47
kandjar` what do you mean?08:47
pheaver see the manpage for git-cherry08:47
'git cherry master bla' will list all commits in bla but not yet in master08:48
see if that list looks reasonable, then just feed that to cherry-pick, i guess08:48
→ anttih joined08:49
kandjar` is there a way to connect both brancheS?08:49
so I could so some kind of rebase08:50
pheaver well, you can try to rebase, and see if it works08:50
:)08:51
→ radarek joined08:51
pheaver or just write a script like (I'm making this up so I don't know if I got it right): for i in `git cherry master bla`; do git cherry-pick $i; done08:51
kandjar` mmm08:54
looks like rebase worked08:54
pheaver sweet08:54
so then, everthing is as it should be?08:55
kandjar` i m checking it08:55
thanks pheaver08:55
oh... last question:08:55
pheaver cool, no problem08:55
kandjar` what s the easiest way to get rid of the other branche? git remote rm ?08:55
pheaver you want to delete the branch with 'git branch -d'08:56
kandjar` just the tracking one08:56
pheaver yeah08:56
kandjar` in gitk it s showing both branches08:56
pheaver yeah just 'git branch -d bla' or whatever it is called08:56
it should happily delete it without complaining, since its history is contained in your master branch08:57
← coderdad left08:57
→ carllerche joined08:58
← ezyang left08:58
→ parolkar joined08:59
→ Weaselweb joined08:59
parolkar While creating tag I get this "-bash: $: command not found"09:00
on Mac OS X leoperd09:00
→ scook0 joined09:00
parolkar any help, Why is it referencing $09:01
← a2800276 left09:11
→ ankit9 joined09:11
→ precision joined09:11
→ yann joined09:14
Ilari parolkar: What command line?09:14
→ pcapriotti joined09:15
tango_ question09:17
→ iulian joined09:17
tango_ I'm converting a project's .cvsignore into .gitignore. should I merge all the .cvsignores in a single .gitignore at the root or would it be better to keep each .gitignore next to the corresponding .cvsignore?09:18
(of course merging the .cvsignores would mean adding the appropriate prefixes09:18
Ilari tango_: Do .cvsignore have the 'appiles in subdirectories' property like '.gitignore' (for entries lacking '/')?09:19
tango_ no idea09:19
→ tvw joined09:20
→ CodeOfficer joined09:21
kandjar` arf.09:22
the tags are local to a repo?09:22
madduck you can push them09:22
Ilari tango_: At least for CVSNT: "The patterns found in .cvsignore are only valid for the directory that contains them, not for any sub-directories.".09:22
kandjar` how?09:22
madduck kandjar`: man git-push09:23
Ilari kandjar`: 'git push <remote> <tag>'?09:23
Gitbot kandjar`: the git-push manpage can be found at http://git.or.cz/man/git-push09:23
kandjar` nice09:24
thanks09:24
:D09:24
ls09:25
ls09:25
oops :) sorry09:25
Ilari tango_: Also, CVS has default patterns to ignore (good candidates for top-level .gitignore): RCS SCCS CVS CVS.adm RCSLOG cvslog.* tags TAGS .make.state .nse_depinfo *~ #* .#* ,* _$* *$ *.old *.bak *.BAK *.orig *.rej .del-* *.a *.olb *.o *.obj *.so *.exe *.Z *.elc *.ln core09:29
madduck checks to see if this is still #git :)09:31
→ ph^ joined09:31
→ hurikhan|Work joined09:33
← daaku left09:39
→ priidu joined09:40
← pheaver left09:44
← anttih left09:54
kandjar` btw, is there any documentation on how to use: git add -i?09:58
→ solydzajs joined10:06
→ exon joined10:06
→ drizzd joined10:07
→ eddyp joined10:13
→ Aikawa joined10:16
→ mcella joined10:18
← aspotashev left10:18
→ ia joined10:18
→ earcar joined10:22
→ name joined10:22
Fla99 so I branched of locally, how do I make sure this branch gets to the remote dir and other people can continue working on it?10:22
← name left10:23
→ fhobia joined10:23
→ name joined10:23
→ parasti joined10:24
→ sergiodj joined10:26
← skrit left10:30
← CodeOfficer left10:30
drizzd Fla99: 'remote dir' <- what's that?10:30
← scientes left10:31
Fla99 my setup is: A local client, B local client and Z is my remote repository10:31
I branched off in A10:31
and want B be able to continue working on that branch10:31
so far i did git push newbranch:origin/head/newbranch10:31
drizzd in B: git fetch A10:31
Fla99 hm..10:32
drizzd pushing into a non-bare repo is slightly dangerous, at least if you push into the checked-out branch10:32
Fla99 that doesn't really work10:32
A is behind NAT and stuff10:32
drizzd faq non-bare10:33
Gitbot drizzd: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare10:33
drizzd you can still do it10:33
but you should read the above first10:33
in any case, your push refspec looks wrong10:33
'origin' is really just an alias to your remote repositories name10:34
it has nothing to do with references10:34
Fla99 so you should never push to a remote repository?10:34
drizzd 11:33 < Gitbot> drizzd: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare10:34
the 'non-bare' is key here10:34
you can _always_ push to bare repos10:35
you have to be careful about pushing to non-bare repos10:35
in any case, your push should look more like "git push origin branch"10:35
Fla99 hm...10:36
and bare meaning?10:36
drizzd a bare repo is just what's under .git without a working directory10:37
Fla99 push origin approaching:refs/heads/approaching is what I did10:37
is that bad?10:37
drizzd ok, sounds reasonable. doesn't work?10:37
Fla99 yes it worked10:37
i can see it now in my remote10:37
now i'm on B10:37
and would like to fetch that10:37
that i don't quite get10:38
drizzd oh, so you pushed to Z?10:38
Fla99 yes10:38
drizzd I thought you were trying to push to B directly10:38
Fla99 ah no no10:38
drizzd ok, now you fetch from Z10:38
Fla99 and the command would be ;) ?10:38
→ charon joined10:38
drizzd git fetch <url-to-Z>10:38
← rubydiamond left10:39
drizzd or "git remote add origin <url-to-Z>; git fetch origin"10:39
Fla99 well that is already configured10:39
→ tango_ joined10:39
drizzd then "git fetch" should do it10:39
Fla99 no10:39
it does not contain the new branch10:39
→ Wyk3d joined10:39
drizzd git branch -r10:40
Fla99 ?10:40
output ""10:40
drizzd lists your 'remote tracking branches'10:40
Fla99 it's null10:40
Wyk3d hello, how can you use mergetool to solve the conflicts that arise from applying patch ?10:40
drizzd and "git fetch" just prints nothing either?10:40
Fla99 * branch HEAD -> FETCH_HEAD10:42
drizzd what does "git config remote.origin.fetch" say?10:43
and check if "git config remote.origin.url" shows the correct url10:43
the former should be something like "+refs/heads/*:refs/remotes/origin/*"10:44
← ankit9 left10:44
→ apeiros_ joined10:44
Fla99 nothing10:45
url seems to be correct10:45
there is clearly a connection10:45
i mean when i push from B that works out nicely10:46
it's just this new branch that I can't get to B somehow10:46
drizzd how did you configure your remote?10:46
Fla99 uh oh. wasn't me :(10:46
drizzd it's because your remote.origin.fetch configuration is missing10:46
Fla99 aha10:46
drizzd someone ****ed it up10:46
Fla99 heh10:47
drizzd you could just do "git remote rm" and "git remote add", that should fix everything up10:47
Fla99 git remote rm10:47
that doesn't do anything dangerous does it?10:48
drizzd or you can "git config remote.origin.fetch "+refs/heads/*:refs/remotes/origin/*"10:48
well, git remote rm removes your remote config, i.e. the remote.origin.url part and it removes your tracking branches10:48
since you still know the url, and you don't have any remote tracking branches, you should be fine10:49
Fla99 hm...10:49
fatal: Invalid refspec '+refs/heads/*:refs/remotes/origin/*10:49
drizzd (the remote tracking branches are the ones shown by git remote -r)10:49
Fla99 ah sorry10:49
typo10:49
ah nice10:49
drizzd Fla99: you can directly edit .git/config, if that's more comfortable10:50
Fla99 hm...10:50
now my local branches don't exist anymore10:51
but I have to do gf checkout origin/branch110:51
and then it says10:51
Note: moving to "origin/approaching" which isn't a local branch10:51
ah10:51
i have the option to create a local branch from it10:51
is that what I should do now?10:52
drizzd very much so10:52
← Wyk3d left10:53
→ vals_ joined10:53
drizzd you should think of the remote tracking refs as "pointers" to the remote repository state. They are only written by "git fetch".10:53
Fla99 :D horray, it seems to be working!10:53
great10:54
really great10:54
i keep messing up with git10:54
drizzd: thanks a lot!10:54
drizzd yw10:54
← tango_ left10:57
Ilari Fla99: All refs in Git are just pointers (also any non-totally-obsolete version keeps log of pointer values) into directed acyclic graph of commits.10:58
→ zumbrunn joined10:58
drizzd Ilari: what if the pointer is removed?10:59
Ilari drizzd: That pointer exists until ref is removed.10:59
drizzd so "git branch -D <branch>", "git gc --prune" should get rid of it for eternity11:00
Ilari drizzd: Except that reflog of HEAD can still reference the commits...11:01
drizzd (get rid of the reflog for <branch> I mean)11:01
← ben_h_ left11:02
Ilari drizzd: Deleting a branch also deletes its reflog...11:02
→ ben_h_ joined11:02
Ilari drizzd: And the 'git branch -D' should also delete the reflog...11:02
→ ankit9 joined11:03
drizzd okay, I thought you claimed the contrary above. my mistake11:03
Ilari drizzd: Actually you don't need 'git gc --prune' unless you want to get rid of the objects. 'git pack-refs' is enough.11:03
→ Lau_of_D` joined11:07
Lau_of_D` Hi everybody - How do I set up a local gitserver on ubuntu, so that I can commit stuff to it from emacs?11:07
Lau_of_D` → Lau_of_DK11:07
Ilari Lau_of_DK: You only need to setup servers if you want to access repository from other computers.11:08
→ askentask joined11:09
askentask when creating and adding to a repo11:09
Ilari Lau_of_DK: And setting up server can be as simple as having SSH open and git installed somewhere that's in $PATH for noninteractive logins.11:09
→ xjjk joined11:09
askentask im in /imaging/ and do git init, then: git add file from /imaging/or the repo?11:09
Lau_of_DK Ilari: So normal webhosting where you just have FTP access is excluded11:10
Ilari Lau_of_DK: Push by FTP is not supported.11:10
drizzd Ilari: you sure? I thought curl covered that11:11
Ilari drizzd: The only 'dumb' push supported is HTTP/WEBDAV.11:11
drizzd oh, sorry nvm.11:11
Lau_of_DK Ilari: Okay, lets say Ive been working on a project in /home/lau/project. And I want to use this as my master git repo. Do I just cd into that directory, git init, and thats that ?11:11
drizzd right, my bad11:11
Ilari The other two supported pushes are git:// (disabled by default for good reasons) and ssh://11:11
Lau_of_DK: 'cd /home/lau/project', 'git init', 'git add .' (or something like that). '.' for add means to add the current directory (recursively).11:12
parasti Lau_of_DK: man gittutorial ...11:13
Gitbot Lau_of_DK: the gittutorial manpage can be found at http://git.or.cz/man/gittutorial11:13
Ilari Lau_of_DK: If you specify directory for 'git add', it adds it recursively.11:13
→ yofel joined11:13
Ilari Lau_of_DK: Well, actually it won't add directories, only files under them (since git doesn't track directories).11:13
Lau_of_DK Ilari: Thanks alot, that'll get me going11:14
parasti: Thanks for the link11:14
Ilari Lau_of_DK: If that project contains stuff like '*.o' or '*~' files (or any such file you don't want to commit), maybe set up '.gitignore' or '.git/info/excludes' before doing the add.11:15
Lau_of_DK Ok11:16
Hmm11:16
Ilari Lau_of_DK: Well, '*.o' is more like stuff for .gitignore and '*~' is more for '.git/info/exclude' (oops, it is 'exclude', not 'excludes').11:16
Lau_of_DK From magit it says "Head: fatal: bad default revision 'HEAD'11:16
Ilari Lau_of_DK: What command says that?11:17
Lau_of_DK magit-status11:18
staged changes: No HEAD commit to compare with (yet)11:18
Ilari Lau_of_DK: Never heard of 'magit-status'...11:18
Lau_of_DK Its the Git-integration for Emacs11:19
Ilari Lau_of_DK: The difference between '.gitignore' and '.git/info/exclude' is that former is tracked, while the latter is not.11:19
Lau_of_DK In practical terms, what difference does that make Ilari ?11:20
Ilari Lau_of_DK: .gitignore is for ignores that are for everyone working with project, exclude is for local ignores (like editor backup files).11:20
Lau_of_DK ah like so11:20
→ Mario_ joined11:21
Mario_ → pygi11:22
Ilari Lau_of_DK: Actually, for editor backup files, better to create '~/.gitignore' (or somesuch) and then do 'git config --global core.excludesfile ~/.gitignore'.11:26
Lau_of_DK: .git/info/excludes is more like for files generated by project and user-specific workflows.11:28
Lau_of_DK Yes I imagine your right. You really dont want to pick up temporary files from any contributor11:28
.gitignore is just newline seperates wildcard-filenames Ilari ?11:28
← pygi left11:31
← askentask left11:32
→ askentask joined11:32
Ilari Lau_of_DK: Yes. For temporaries, you can use parts of CVS default ignore list: '#*' (thats bit tricky to match, but '[#]*' seems to work), '.#*', '*~', ',*', '*.old', '*.bak', '*.BAK', '*$', '_*$'. You also might want to ignore '*.orig', '*.rej', '*.so', '*.a' and '*.o'. Also '*.exe', '*.dll' and '*.obj' if project is (also) for Windows.11:34
Arachnid Is it possible to revert just one file to the currently checked-in state?11:35
Ilari Lau_of_DK: '*.so', '*.a' and '*.o' are the unix equivalents for '*.dll', '*.lib' and '*.obj'. That list was missing '*.lib' from Windows section.11:35
Arachnid: 'git checkout HEAD -- file'.11:35
Arachnid thanks :)11:35
→ asdf joined11:36
asdf → Guest9117411:37
← askentask left11:37
Ilari Lau_of_DK: Also, '/' is special. If present, it modifies the way that pattern is matched.11:37
Lau_of_DK Ilari: Alright, thanks alot, youve been instrumental11:37
→ _graham_ joined11:40
Ilari Lau_of_DK: Patterns that contain '/' are only valid in directory that contains the .gitignore (exclude behaves like its at root level), unlike patterns that do not (they get appiled recursively). '/' at beginning also restricts pattern to match only in current directory.11:42
→ akitada joined11:43
Ilari Lau_of_DK: Also, patterns beginning with '!' un-ignore files matching the pattern. Also, you can force adding ignored file (tracked files are not affected by .gitignore or excludes).11:43
→ kumbayo joined11:44
← parasti left11:45
→ nud joined11:45
Ilari Lau_of_DK: See man gitignore11:46
Gitbot Lau_of_DK: the gitignore manpage can be found at http://git.or.cz/man/gitignore11:46
Lau_of_DK Ilari: You wouldnt happend to be the auther of the .gitignore parser would you? :)11:56
← radarek left11:56
← ph^ left12:04
Ilari Lau_of_DK: Nope... :-)12:04
→ ph^ joined12:06
→ DrFrasierCrane joined12:07
→ kalila joined12:11
kalila hello - what are my options for settings up a git server for a group of people? (50+). the size of the repos is colossal, (subversion). There are many options (raw git? git over http? direct in place?) I don't know what to pick12:12
what does the kernel use?12:12
wereHamster you can pull the kernel via http or git and developers push via git+ssh/ssh12:13
kalila they're all operating on the same files, or is there some replication?12:14
wereHamster git:// is faster then http://, the downside is that you have to open another port in your firewall for it to work.12:14
same files12:14
Ilari kalila: 'repos' in plural and group of people? Sounds like you should look at Gitosis...12:15
kalila: Authenticated fetching/pulling/cloning can also be done using ssh://12:15
kalila: ssh:// is as fast as git://12:16
intripoon hi12:16
kalila well, it's one repo that we slap everything to12:16
I'll be looking into that, thanks12:16
intripoon How do can I extract the version 3 commits ago from my repository?12:17
kalila -r head~3 ?12:17
intripoon I don't know12:17
→ naeu joined12:17
intripoon how do I get a list of things I can -r ?12:17
Ilari intripoon: Depends on meaning of extract... Anyway, the SHA expression for that commit is 'HEAD~3'...12:18
← carllerche left12:18
→ Sho_ joined12:18
intripoon Unknown option -r12:18
I'm using git version 1.5.4.2.1161.g1a6f012:19
Ilari intripoon: What are you trying to do anyway?12:19
kalila thanks guys for the tips12:19
intripoon I just want to look into some files and see what has changed12:19
Ilari intripoon: 'git diff HEAD~3'?12:19
bremner intripoon: using git diff maybe?12:19
intripoon I'ld prefer winmerge ...12:20
Ilari intripoon: That's diff between version three commits ago and current working tree... If you want to diff the latest commit, use 'git diff HEAD~3 HEAD'.12:20
← kalila left12:20
intripoon between all files?12:21
← mcella left12:21
intripoon "git diff HEAD~3" gives me "fatal: ambiguous argument 'HEAD~3': unknown revision or path not in the working tree"12:22
Ilari intripoon: You can also add '-- file1 file2' (or so) to limit the comparision to only some files or paths.12:22
intripoon I don't like reading diff anyways12:22
Can't I get the working tree from 3 commits ago?12:23
Ilari intripoon: Does the branch you are on have enough commits for "three commits ago" to make any sense?12:23
bremner intripoon: sure, use git checkout12:23
intripoon I only have a master branch12:23
Ilari intripoon: You could extract a snapshot of it: 'git archive HEAD~3'.12:23
intripoon and commit to oit all the time12:23
bremner Ilari: won't 'git checkout HEAD~3' work?12:24
Ilari bremner: Yes. But that would change the working tree files...12:24
intripoon git archive HEAD~3 outputs a lot of stuff12:24
Ilari intripoon: That's tar archive of that revision...12:25
→ moccuo joined12:25
Ilari intripoon: Pipe it to some file or to tar process running in suitable directory...12:25
bremner intripoon: if you don't mind changing your working tree (i.e. if that makes sense to you), then you could use checkout12:26
← fhobia left12:26
← earcar left12:26
intripoon I could clone and then checkout12:26
what can I put there instead of HEAD~3 ?12:26
are the commits named somehow?12:27
bremner intripoon: you could, but since you could always checkout your master branch again, it is not really required12:27
intripoon can I get a list of commit's names?12:27
bremner intripoon: git log12:27
Ilari intripoon: If you clone a temporary repository, use 'clone -s' (be aware that the new repo needs the old to be present) to make the clone operation as fast as possible.12:28
← ajonat_ left12:29
bremner is reminded of Knuth's statement on premature optimization being the root of all evil :-)12:29
intripoon git log gives me a list of my commits, and now I use this long number behind the word commit instead of HEAD~3 ?12:30
bremner intripoon: yup. Cut and past?12:30
paste even12:30
intripoon let me see12:30
bremner intripoon: actually you only need the first few characters (5?)12:30
Ilari intripoon: 'winmerge' sounds bit windowish, so you can get .zip of revision using something like 'git archive --format=zip HEAD~3 >version-three-commits-ago.zip' (or something like that).12:31
intripoon: Actually, how many first characters you need mostly depends on size of repository... Absolute minimum is four. Large repositories could require 10 or more...12:31
intripoon it looks like this commit d416c32aee226e774b51dcbba500039dfb89480112:32
bremner yup, so git checkout d416c32aee226e12:33
intripoon and yes I'm on windows12:33
Ilari intripoon: Commit IDs are also valid SHA expressions to specify a revision...12:33
← Sho_ left12:35
→ earcar joined12:37
→ ashleyw_ joined12:38
Ilari intripoon: SHA expression consists of base part, followed by optional operators. Base part can be object ID, unique abbrevation (of at least four hexes) of it, name of branch, name of tag, name of ref without initial 'refs/' or fully qualified name of ref. Operators include '^' (first parent), '^n' (nth parent), '~n' (first parent iterated n times), '^0' (commit itself), '^{}' (peel tags), '^{type}' (peel until object of 'type' is reached), '12:41
Oops... That might be too long... Where it cuts off (it doesn't show cutoffs here)?12:41
← quoin left12:41
→ dhutty joined12:42
← earcar left12:44
Guest91174 http://github.com/xxxx/shapefile-generation/tree12:45
intripoon wow12:45
Guest91174 now, i followed those steps but when i tryto: git push origin master12:46
Ilari It should end with "get ID of object corresponding to path in given revision)."12:47
Guest91174 i get: src refspec oes not match any12:47
Ilari Guest91174: Does 'git log master' give anything?12:47
Guest91174 remote end hung up unexpectadly12:47
intripoon ok, that helped me a lot12:48
thx guys12:48
→ Yuuhi joined12:49
Guest91174 → asken12:50
→ niki joined12:50
asken do you commit inside the repo?12:52
→ Sho_ joined12:52
← ashleyw left12:52
asken anyone good with github?12:52
im getting really frustrated here12:53
Ilari asken: 'git commit' makes commit in local repo.12:53
asken: And you need at least one commit to make a push...12:53
asken i did one commit12:53
what i dont egt it is inside the repo bvs outisde12:53
← Fla99 left12:53
Ilari asken: Does 'git log master' show that commit?12:53
asken i am in ~/imaging/ and in there i have the repo12:53
ambiguous statement master12:54
Ilari asken: Also, what git version?12:54
asken how to chekc?12:54
Ilari asken: 'git --version'.12:55
asken 1.6.0.212:55
newest?12:55
Ilari asken: That's definitely new enough...12:55
← priidu left12:56
asken i tried: git log master12:56
→ bdiego joined12:56
asken ambiguos staement master12:56
Ilari asken: Well, What does 'git branch' give? Nothing?12:57
asken: "fatal: ambiguous argument 'master': unknown revision or path not in the working tree."?12:57
asken yes12:57
→ flaguy joined12:59
→ ciskje joined12:59
Ilari asken: Then there probably are no commits (they are normally created using 'git commit' or 'git commit -a' (if you don't understand the index)).12:59
asken i did12:59
look im ~/imaging/ where i have shapegen.clj13:00
now i want to make a repo otu of this13:00
how?13:00
Ilari asken: 'git add shapegen.cij', 'git commit'. It should pop up editor window asking for commit message.13:00
asken i did: git init; git add .; git commit -m "firsttry";13:00
→ rubbish joined13:02
Ilari asken: Did the commit succeed? It should have printed something like '[master (root-commit)]: created yyyyyyy: "firsttry"'...13:02
asken donkey donkey me13:02
i had another dir with the repo13:02
wow13:03
worked its on github13:03
so cool13:03
github is awesome13:03
→ flice joined13:03
flice hi13:03
asken make sharing and collab so eays i think it will13:03
Ilari asken: The more trivial the mistake, the harder it is usually to find...13:04
flice I'm trying to use .git/info/exclude, but it doesn't work for me for some reason. the goal is to ignore a directory that is present upstream. this means not to get upstream commits in that directory, etc13:04
the pattern I'm using is mydir/**13:04
→ ciskje_ joined13:04
flice also tried mydir/13:05
albertito flice: /mydir?13:05
asken Ilari: yes that is true13:05
albertito flice: there's also .gitignore, if you want to have that ignore pattern inside the repository13:05
flice: (man gitignore has the documentation)13:06
Gitbot flice: the gitignore manpage can be found at http://git.or.cz/man/gitignore13:06
← Flathead left13:06
janl_ flice: ignores are only for untracked files13:06
Ilari flice: Ignore patterns don't affect tracked files...13:06
flice albertito: the point is that I don't. it should be local13:06
so is there no way I can just ignore part of the tree that is present in the repository?13:06
albertito flice: use .git/info/exclude then =)13:07
flice albertito: that's what I'm using13:07
albertito flice: if it's already in the repository, no13:07
flice what about "subprojects" or something? with 149 commands there should be a way13:07
wereHamster flice: explain why you are trying to ignore that subdirectory13:10
bremner sounds like assume-unchanged might help13:11
→ rubydiamond joined13:11
flice wereHamster: the upstream repository is shared between a small team, and it has 2 unrelated projects in it. I don't want to create another repo for technical reasons, so I'd prefer a way to separate between the projects in a single repo13:11
← name left13:11
bremner forgets the exact syntax for setting assume-unchanged flag13:11
flice bremner: will investigate that, thanks13:12
wereHamster flice: so it's simply that you are not interested in the directory?13:12
→ komplex joined13:12
flice wereHamster: exactly13:12
wereHamster no other technical reasons?13:12
janl_ flice: you can can unrelated branches that doest share any files, google for git empty branch13:12
flice janl_: can I have 2 branches simultaneously checked out in separate directories, but inside one repo?13:13
janl_ flice: not in the same worktree no13:14
flice: but you can ofcourse have two clones of the repo..with one branch checked out in each13:14
flice janl_: I oversimplified the problem a bit. actually, I need to work on both projects simultaneously, while other people are only interested in one of the projects. so making 2 branches would require me to switch every time13:15
janl_: that's an idea worth thinking of. though I'd prefer something more simple13:15
janl_ flice: ah, yes, probably not the best solution then13:15
vals_ → tango_13:16
flice may assume-unchanged help?13:17
→ wrobbie joined13:17
bremner flice: I have only read about it, see man git-update-index13:19
Gitbot flice: the git-update-index manpage can be found at http://git.or.cz/man/git-update-index13:19
← ciskje left13:19
Ilari janl_: There's 'git-new-worktree' in contrib...13:20
asken how do i remove a file from a project?13:20
Ilari janl_: With it, be careful not to check out the same branch twice (or you can get very confusing results).13:20
asken: 'git rm file' and then commit the removal.13:21
bremner flice: hmm, I think upstream changes will still trouble you13:21
Ilari asken: Or 'git rm --cached file' if you don't want it removed from disk...13:21
flice bremner: yep, just checked it13:21
bremner flice: basically you are running up against the git doesn't yet support subtree checkouts problem13:22
flice probably two branches in separate local repositories is the way to go13:22
← naeu left13:22
flice bremner: yes, looks like that13:22
weird, since even cvs supports it13:22
what about tags? perhaps I can tag both projects with tag A and then project 1 with tag B and project 2 with tag C. then I personally would check out A and other people would check out either B or C13:23
Ilari flice: It is possible for two working trees to share a repo...13:24
flice Ilari: that's a back up option janl_ suggested13:24
asken wow github is very cool13:24
flice but then (with tags) I think each time someone adds a file to the project, they'll need to tag it twice13:25
→ ashleyw joined13:25
Ilari flice: Files are not tagged, revisions are.13:25
← rubydiamond left13:25
flice Ilari: no sticky tags?13:26
bremner flice: why can't people just ignore the directory they are not interested in?13:29
← ashleyw_ left13:29
Ilari flice: Git only knows three types of tags: Lightweight tags which are just pointers to commit DAG, annotated tags which also have object containing description of tag, and signed tags, which are annotated tags that also contain GPG signature.13:29
→ kreaturr joined13:29
bremner flice: another option, which is kindof gross, is use git-filter-branch to kill the directory you hate, and then only update the parent repo via patches13:30
janl_ ugh :)13:30
bremner is master of understatement "kindof gross", throw away half of git13:30
Ilari Well, actually any type of object, not just commit can be tagged (yes, you can tag a annotated/signed tag). But usually only commits are tagged.13:30
← dhutty left13:31
← ankit9 left13:31
drizzd flice: you could have to repos, one for the directory you hate, and one for the rest. you can combine them using submodules13:32
s/to repos/two repos/13:32
bremner if he could have two repos, we would be done13:33
asken i thoguht git was complicated? it seems very easy to use13:33
→ ankit9 joined13:34
bremner asken: that's not what you said last night :-)13:34
asken yes but i got it working now and finally i got the basics clicking13:34
drizzd actually, if you just 'git rm' the subdir in one branch and re-add it during merge with upstream. similarly, you merge upstream into your branch to synch. rerere should take care of the removing and re-adding13:35
→ dhutty joined13:37
dhutty is this an appropriate venue for git-newbie questions? or is this more of a dev channel?13:38
bremner dhutty: newbs R us13:38
dhutty cool. bbiab once I get my thoughts sorted enough to ask something sensibly:)13:38
Lau_of_DK dhutty: we are the noobs13:38
→ gitte joined13:41
← kumbayo left13:43
→ Bass10 joined13:46
→ kumbayo joined13:47
→ naeu joined13:51
→ hidek joined13:54
→ snitko joined13:55
→ sgarza joined14:00
→ vmlinz joined14:03
← sgarza left14:05
← kandjar` left14:06
← ben_h_ left14:07
← kumbayo left14:07
← ankit9 left14:09
→ coderdad joined14:11
→ kalila joined14:12
kalila Can gitosis accept all users of its server14:12
without the need to add them by hand one by one14:12
← yofel left14:18
← coderdad left14:18
gitte kalila: my understanding was that gitosis separates the user management completely from the host's user management.14:19
Disclaimer: I am by no means an expert in Gitosis.14:19
→ drizzd_ joined14:20
→ Zimm_ joined14:23
→ agile joined14:23
Run Is there a way to see the diff's that will be pushed to remote just prior to doing 'git push'?14:25
kalila I'd like to have a git server that authenticates agains the host's users (which fetches it from LDAP to start with).14:25
whatever possible: gitosis, http, etc14:26
← niki left14:27
← drizzd left14:27
→ samjam joined14:28
→ drizzd joined14:29
gitte Run: try to run "git ls-remote" with that remote... you'll see what the remote's idea of the current tips of the branches is.14:30
→ cedricv joined14:31
→ brett_h joined14:32
→ cyclist joined14:32
← parolkar left14:33
← agile left14:33
← naeu left14:33
Ilari kalila: SSH?14:36
→ niki joined14:37
→ cedric_ joined14:38
→ kumbayo joined14:38
Ilari kalila: Otherwise I think you need to hack Git source... Perhaps the easiest modification in client side would be to support connecting to Unix domain sockets. Then the server could get the PID, UID and GID of process (and from PID, you can get stuff like {R,E,S,FS}[GU]ID, supplementary groups, Security context, etc)...14:40
kalila Ilari: thanks14:40
I'll try the straightforward ssh14:40
← drizzd_ left14:41
Ilari kalila: The advantage of going through unix domain sockets is that authentication would be very easy for users (assuming they have apporiately modified Git).14:41
→ coderdad joined14:42
Ilari kalila: Hmm... Maybe it could be possible to do by abusing GIT_PROXY_COMMAND. That way one can use unmodified git binary.14:43
→ ^Chechen^ joined14:43
← ^Chechen^ left14:44
→ yofel joined14:44
→ kukks joined14:45
→ Voker57 joined14:45
→ aspotashev joined14:47
← cedricv left14:47
→ lilit joined14:48
→ Kangou joined14:48
→ earcar joined14:48
→ d0k joined14:49
← hurikhan|Work left14:51
← cyclist left14:51
→ hurikhan|Work joined14:52
← cedric_ left14:52
asken → blenket14:55
→ tanek joined14:56
→ xkr47 joined14:56
← snitko left14:56
← Lau_of_DK left14:57
← coderdad left14:58
← kalila left14:58
→ pygi joined14:59
→ ashleyw_ joined15:00
→ coderdad joined15:00
→ kek joined15:00
Ilari Kangou: Problem is that with unix domain sockets, you need two paths, first for designating the socket to connect to and second to specify the repository...15:00
→ tritz joined15:01
dhutty one of the reasons I moved to svk from svn was so that I wouldn't have .svn files and the like littered all over my working copy hierarchy. Now I'm trying git and when I do git init, I get .git files in the pwd. Am I misunderstanding something?15:02
Run dhutty: Yes... because those files never have bothered me. What do you want to do that those directories are in your way?15:04
← Sho_ left15:05
blenket i can have several project son github right?15:05
dhutty Run: find(1), filesystem operations in general15:05
Voker57 dhutty: there's only one .git in root directory of working tree15:06
← kumbayo left15:06
dhutty well that's better than every directory like svn:)15:07
Run dhutty: You can tell find to ignore .git, but in general, you'd first export the tree (with git-archive) in order to really get rid of those administration files.15:07
dhutty if y'all think it's not a problem, I'll just use it and see what happens15:07
→ priidu joined15:07
Run In most cases there is a way to ignore the dir (ie, a diff).15:08
dhutty I know I _can_ tell it to exclude certain things, but I don't want to _have_ to.15:08
Run Well, if you run 'find' in the root then there is no way for find to know what you want unless you tell it to.15:09
→ JensB joined15:09
JensB Hi15:09
dhutty I'm well aware of that.15:09
→ LiamH joined15:09
← akitada left15:10
Run The only way to officially get rid of .git is with git-archive afaik15:10
jast no15:10
Run Just saying that for most cases that is not needed.15:10
jast that was tla15:10
→ Sigma joined15:10
dhutty that's why I don't want anything there that isn't a file that I'm working on.15:10
Run tla?15:10
JensB I created a Git repository 32 commits ago. Now I want to push all these commits into an (empty, so far) SVN repository. "git svn init" seems to only work with empty git repos ... how do I push to SVN?15:11
jast the original GNU Arch15:11
dhutty Run: ok, thanks for the info15:11
gitte JensB: I fear that you'll have to "git svn clone -s" that repository first.15:12
Run Sounds like he wants to switch to svn instead of git.15:12
→ radarek joined15:12
JensB Run15:12
no.15:12
gitte JensB: then you can fetch that into the other, and rebase your commits.15:12
Run: probably just wants to use Google Code or something.15:13
JensB gitte, neither15:13
→ Beket joined15:14
jast svn init should work in any repo15:14
JensB jast, ok .. will try that (after making a backup)15:14
jast git svn init, i mean15:15
JensB gitte, I prefer using Git locally because I'm offline often, and with Git I can branch and merge away without being connected to the uplink svn.15:15
jast, thanks15:15
← ashleyw left15:16
jast the thing is that you can't create the svn repository base layout this way15:16
Ilari JensB: There's 'git2svn'. Don't know if it can be used in this case...15:17
JensB: Also, I have once hacked set of scripts that could send commits to SVN (with proper layout, but no copies/renames)...15:18
jast i suggest you use normal svn to create the branches/tags/trunk dirs, then git svn init, followed by git svn rebase15:19
Ilari JensB: Well, actually those scripts where written with "it only has to work for me" mentality (and it really shows in the code)...15:19
← coderdad left15:19
→ ankit9 joined15:19
→ cedricv joined15:19
← apeiros_ left15:20
→ apeiros_ joined15:20
JensB tried "git svn init svn+ssh://......." in my repo and everything seems to still be there15:20
jast well, what did you expect? (and don't you want to use -s? See also my suggestion above regarding that)15:22
JensB jast, the doc says "git svn init initializes AN EMPTY GIT REPOSITORY with ..." <- I was half expecting mine to be overwritten.15:23
← hurikhan|Work left15:25
jast nah, it just says that because you usually don't use it when you've already committed stuff15:25
all it does is set up a bit of metadata and config settings15:25
JensB ok15:25
now "git svn dcommit" says "Can't call method "full_url" on an undefined value at /usr/local/git/libexec/git-core/git-svn line 425."15:25
jast dcommit won't work yet15:26
JensB ok15:26
how do I push all my commits into the remote SVN?15:26
jast it only works if there is *something* in the svn repository, i believe15:26
that's why i suggested what i did15:26
← wrobbie left15:28
← cedricv left15:28
→ rmh3093 joined15:33
← vmlinz left15:34
→ LuboC joined15:34
→ muthu_ joined15:37
blenket i created a branch experimental15:38
← moccuo left15:38
blenket now i want to adda file to it, how? git add fiel normall but that adds to master right?15:39
→ drewr joined15:40
Voker57 how did you create it?15:40
if you did `git branch experimental`, you need to `git checkout experimental`15:41
`git checkout -b experimental` would create a branch and checkout it15:41
→ coderdad joined15:42
→ felipec joined15:43
→ tokkee joined15:43
blenket so if i add a file to master and then do experimental, and then i change the file...15:44
a file cant be "owned" by a branch?15:45
← GNUix left15:47
JensB jast, in the docs it says that "git svn clone -s" is the same as "git svn init -s && git svn fetch". I did both, I still get the error above.15:47
Voker57 blenket: it can15:48
→ moccuo joined15:48
Voker57 if you add it and commit15:48
jast JensB, yes, as i said, it doesn't work without pre-initing the subversion repository using the svn client.15:54
and forget about git svn clone, that actually creates a new repository15:54
← d0k left15:55
→ d0k joined15:56
→ gitte_ joined15:56
← gitte left15:57
gitte_ → gitte15:57
← d0k left15:59
JensB jast, I'm sorry if this sounds dumb, but what would I have to do? the svn repo isn't empty but I wouldn't care about overwriting it.16:01
→ d0k joined16:01
JensB since all my commits are in my local git repo16:01
jast oh, it isn't16:01
← tango_ left16:02
JensB thanks16:04
jast hmm.16:04
JensB so - is there any way I can create a SVN repo and clone my Git repo into the (new) SVN repo with all history?16:04
jast what's in the svn repo? random commits with code in them?16:04
JensB or overwrite the existing SVN repo16:04
jast: a single commit with a default Ruby on Rails app, no content.16:05
ie. can be overwritten.16:05
jast, my git repo is extremely simple: totally linear, no branches, no merges, one developer.16:08
I'd just like to populate my bosses SVN repo with my work.16:08
jast JensB, so do you want the typical branches/tags/trunk structure in the svn repo?16:09
JensB jast, if possible yes, but not strictly necessary. if there is a way to mirror Git branches in SVN, would be nice, but not strictly necessary.16:10
← coderdad left16:10
jast mirroring branches to svn is a partially manual process but it's possible16:10
bremner Is rebasing a range of commits including a merge (i.e. squashing a commit before the merge and one after) going to be problematic?16:10
JensB maybe the following works: a) Create a new Git repo via "git svn init -s"; b) git svn fetch; c) pull/import/clone(?) first commit of my work Git repo; d) work out conflicts; e) pull/import/clone(?) rest of my Git commits into new repo; f) git svn dcommit16:11
JensB is looking at tailor16:12
→ mcella joined16:13
jast bremner, rebase destroys merge commits16:15
(iirc)16:15
BeerSerc hey there. I am getting the following tar error when trying to make install install-doc git from it's git repo: http://rafb.net/p/F7SveO31.html16:15
bremner jast: ok, I made a test branch. learning by doing...16:15
jast hmm, rebase -i has -p for preserving merges16:16
→ coderdad joined16:16
jast JensB, i'll be gone for a while, perhaps we can work this out afterwards16:17
bremner well, without -p was no good. Conflicts when I picked everything.16:17
dhutty is git able to keep file permissions? e.g. If I have a file that is 400 when committed will it be 400 when checked out on another machine?16:18
Voker57 yes16:18
jast git only looks at the user permissions16:19
all it stores internally is 644 and 755; the umask is applied when actually checking out files16:19
dhutty ok16:19
→ ashleyw joined16:21
JensB jast, thanks - I'll keep trying16:21
→ jackdempsey joined16:24
JensB ok, let's try a different approach16:26
← ashleyw_ left16:27
bremner hmm, after git rebase -i -p and leaving all commits as pick, some files became untracked. Could this be a bug? git 1.5.6.516:28
→ ajonat joined16:31
← mcella left16:34
gitte Yep.16:34
→ softdrink joined16:34
gitte It could even be fixed in the latest Git ;-)16:34
BeerSerc no ideas why that bash part in make install fails here?16:40
← flaguy left16:43
→ kalila joined16:43
kalila "git checkout" is not at all similar to svn checkout, right?16:43
Voker57 well kinda16:44
→ anttih joined16:44
Voker57 it checks out fiels too16:44
kalila I don't get the difference16:44
Voker57 they _are_ similar16:45
the only difference is SVN checks out files from remote repo and git -- from local one16:45
well and git checkout does not create working tree16:46
kalila can't do: git co username@server:<path/to/dir> ? (like clone)16:46
Voker57 no16:47
← blenket left16:48
← coderdad left16:50
→ mosno joined16:51
← tanek left16:52
Kangou → Kangou_16:53
Kangou_ → Kangou16:55
→ tanek joined16:56
jast kalila, in git, checkout is used mainly for switching branches in local repositories. the only real counterpart to svn co is git clone, with the difference that it fetches all history and all branches at the same time.16:58
bremner gitte: Thanks, I'll give it a try. I just wondered if I expecting too much (with merge commits and rebase).16:58
JensB jast, I think I have found a solution to my problem: https://kerneltrap.org/mailarchive/git/2008/10/25/3808004/thread16:58
jast, haven't got it to work yet though16:59
gitte bremner: basically, the -p thing is so rarely used that it is only barely tested.16:59
bremner: also, keep in mind that there is _no way_ to recreate the merges perfectly.16:59
← radarek left17:02
→ coderdad joined17:06
→ tim_ema joined17:07
← tim_ema left17:07
→ hurikhan|Work joined17:07
← coderdad left17:08
→ EmilMedve joined17:09
→ coderdad joined17:09
← coderdad left17:11
→ andreas joined17:22
andreas TML arer u there?17:23
→ bobesponja joined17:25
→ Niamor joined17:26
→ j_engelh joined17:30
← mosno left17:31
← rubbish left17:33
← flice left17:34
→ rubbish joined17:35
→ chris2 joined17:36
→ ashleyw_ joined17:36
andreas TML hi17:36
← nud left17:37
Run Trying to remove a topgit branch :/17:42
What a MESS!17:42
How can I recover from:17:42
hikaru:master:slviewer>tg summary17:43
fatal: ambiguous argument 'debian/Allow_CXXFLAGS_to_specifiy_the_debug_flag': unknown revision or path not in the working tree.17:43
Use '--' to separate paths from revisions17:43
fatal: bad revision '^debian/Allow_CXXFLAGS_to_specifiy_the_debug_flag'17:43
fatal: Not a valid object name debian/Allow_CXXFLAGS_to_specifiy_the_debug_flag:.topdeps17:43
fatal: bad revision '^debian/Allow_CXXFLAGS_to_specifiy_the_debug_flag'17:43
fatal: ambiguous argument 'debian/Allow_CXXFLAGS_to_specifiy_the_debug_flag': unknown revision or path not in the working tree.17:43
Use '--' to separate paths from revisions17:43
fatal: Not a valid object name debian/Allow_CXXFLAGS_to_specifiy_the_debug_flag:.topmsg17:43
0r debian/Allow_CXXFLAGS_to_specifiy_the_debug_flag17:43
j_engelh enough of the debian problems :p17:43
Run that is just the branch name17:43
← ciskje_ left17:44
jast yeah, i guess tg didn't like that you removed the branch yourself17:44
unfortunately i can't tell you what to do now because i don't know anything about tg17:44
by the way, please consider pasting longer passages of text to http://git.pastebin.com/ rather than into the channel17:45
Run sorry17:45
frustration :)17:45
jast send me a lot of money and i'll forgive you17:46
j_engelh don't forget me!17:46
jast i think j_engelh has a highlight configured on "money"17:46
→ ChugChug joined17:48
ChugChug So, I wanted to try something out and created a branch for it. After a few checkins, I realized that what I was trying to do was stupid. Now I just want to delete the branch. How do I do what without merging it back in?17:50
j_engelh delete it.17:50
← kalila left17:50
bremner ChugChug: man git-branch17:51
Gitbot ChugChug: the git-branch manpage can be found at http://git.or.cz/man/git-branch17:51
← ashleyw left17:52
ChugChug hehe... In hindsight, was an embarrassing question. Thanks much!17:52
bremner ChugChug: no prob, I just like to make GitBot work17:53
ChugChug :)17:53
Run jast: Ok, we fixed it - heh, I think. At the errors are gone.17:54
JensB jast, I think I have found a solution to my problem: https://kerneltrap.org/mailarchive/git/2008/10/25/3808004/thread17:54
jast, worked fine - but the rebase linarized my history completely ... *shrug*17:54
jast that's what rebase does17:55
→ patmaddox joined17:55
JensB well, no big pain17:55
andreas I pasted my new files to my git directory on my mashine and want to paste them to my github directory to but when i used git push it only send the readme up does someone know how to deal with that?17:55
jast so do things work now?17:55
Run jast: If I do a 'tg summary' I see some 'L's, meaning that my branches are 'ahead' of the origin... So, I should do a 'git push' to sync it. However, I want to know why and what is different. What command would show me a diff of what will be pushed on a given branch when I do 'git push'?17:55
jast andreas, have you committed the changes? push only transmits commits.17:55
andreas jast I'm pretty new so i dopnt know how17:56
jast Run, typically, git log origin/foo..foo (show commits that are in the latter but not in the former)17:56
Run kewl17:56
← rubbish left17:57
→ stuffcorpse joined17:57
jast andreas, i suggest that you read a git tutorial then, or the git manual. they explain how things work. frankly, it's a bit much to explain here.17:57
sure, i could tell you which commands to use but that helps little if you don't understand the reasoning17:57
andreas ok thanks17:57
jast feel free to ask if anything in the docs is unclear17:58
bremner gitte: FYI, the bug I saw with rebase -i -p is indeed gone in 1.6.0.417:58
gitte bremner: thanks for the report.17:58
bremner now it says17:59
"Refusing to squash a merge" when I try to squash two commits on opposite sides of (several) merges, but I guess that is just life18:00
gitte Well, you have to move the commits around that you want to squash, so that one comes directly after the other.18:01
And then you have to mark the _second_ as "squash".18:01
(You do not want to squash the first into the previous commit)18:01
jast squash means: squash this commit into the one on the previous line18:01
bremner gitte, jast, right, I did move the commit I wanted to squash forward to right behind the first and mark it as s18:03
gitte If Git still complains, then one of the two commits _is_ a merge.18:04
bremner checking...18:04
gitte You can still squash it if you want, by marking the second as "e", and then playing games with git-reset and commit-tree, _but_ you probably do not want an evil merge, right?18:05
→ ashleyw joined18:05
bremner nah, I'm not that motivated.18:05
gitte What I mean: even if you were motivated, you probably do not want to squash anything into a merge (or squash a merge into anything).18:06
← ajonat left18:06
bremner gitte: ahh, I see. No, I want to leave the merges alone as much as possible18:06
andreas jast my git showed this wheni typed git commit "18:07
bremner niether commit looks like a merge in gitk.18:07
andreas Connection closed by 65.74.177.12918:07
fatal: The remote end hung up unexpectedly18:07
vmiklos gitte: have you got the fast-export mail i sent? (not bugging, just checking if i hit a bug or i'm just not understanding something)18:07
gitte vmiklos: I got it, but felt too lazy today to do anything about it.18:07
vmiklos ok :)18:07
gitte vmiklos: first thing I'd need to do is turn that into a test script, anyway.18:07
→ skrit joined18:07
→ CodeOfficer joined18:08
← sergiodj left18:08
vmiklos gitte: you mean a new testcase to t9301-fast-export.sh?18:08
gitte probably that, yes.18:08
vmiklos ok, i can do that.18:08
← DrFrasierCrane left18:12
→ kek_ joined18:12
← ashleyw_ left18:13
→ rubbish joined18:13
andreas jast ok i got now this vim like windowe and editted it but i dont know how to get on18:14
jast what you're doing is writing a commit message. finish that, save and quit.18:15
→ dreiss joined18:15
← JensB left18:19
← rmh3093 left18:23
← kek left18:24
→ ajonat joined18:25
→ rmh3093 joined18:25
← chris2 left18:26
→ chris2 joined18:27
→ naeu joined18:27
→ x_or joined18:28
← Niamor left18:32
← LuboC left18:37
← moccuo left18:38
← gitte left18:39
→ gitte joined18:39
→ priidu_ joined18:40
← priidu left18:40
← priidu_ left18:40
→ priidu joined18:40
→ moccuo joined18:40
→ coderdad joined18:41
← rubbish left18:48
→ mcella joined18:49
→ robinr joined18:50
andreas jast IM a newbie i dont now how to save and quit it18:52
TML hear me?18:52
drizzd andreas: Alt+F418:53
j_engelh IHNIWYM18:53
andreas thart closed the terminal18:54
j_engelh you have to click on your IM window first18:54
vmiklos j_engelh: huh, failed to resolve that abbreviation :)18:54
drizzd I have no idea what you mean18:54
j_engelh vmiklos: same with TML18:54
andreas drizzed my terminal cled18:55
vmiklos ah :)18:55
j_engelh hit Ctrl-Alt-Backspace twice18:55
drizzd andreas: you wanted to quit 'it', Alt+F4 quitted 'it'. Why are you complaining?18:55
vmiklos j_engelh: tml is a nick :)18:55
andreas drizzd i wanted rto save and quit it it was my commit tzo git hub18:56
j_engelh (don't drink and IRC!)18:56
→ Niamor joined18:56
j_engelh 118:56
drizzd andreas: ok, I'm going to presume that you did not configure your $EDITOR18:57
so by default the 'vi' editor is started to edit the commit message18:57
→ Jacolyte joined18:57
drizzd if that's not what you want, you should set "git config --global core.editor my-favorite-editor"18:58
or you can also set the environment variables GIT_EDITOR, VISUAL, or EDITOR to the same effect18:58
← ashleyw left19:00
→ SRabbelier joined19:02
← bdiego left19:03
→ name joined19:07
j_engelh EDITOR is best19:07
← kek_ left19:07
→ ashleyw joined19:09
andreas well I typed git commit again and editted the lines and wrote under it git add and the names of the files that should be added what now??19:10
thiago_home once you've committed, it's done19:11
Ilari andreas: Save and Quit the editor?19:12
andreas aha and then how tio push it to the git repo master br<anch19:13
thiago_home git push repo master19:13
Ilari andreas: You only need to specify what branch to push if you want to push new branches that don't exist in upstream yet.19:13
→ schierbeck joined19:13
gitte andreas: which tutorials did you read?19:14
Ilari andreas: So, after 'git push repo master', 'git push repo' will push master.19:14
→ harinath joined19:14
schierbeck is there a way to programmatically remove a submodule?19:14
gitte schierbeck: do you mean from the index, from the working directory, or both?19:15
schierbeck gitte: both19:15
andreas I 'M a fuckking newbie pleasde go easy on me19:15
schierbeck i'm writing a plugin management api for rails19:16
andreas how to save and quit a default editor like vi?19:16
gitte schierbeck: the first can be done by "git rm --cached <submodule>" I guess...19:16
→ DrFrasierCrane joined19:17
→ spearce joined19:17
gitte schierbeck: the 2nd is done by "rm -rf <subdirectory>". But beware. Anything _unpushed_ will be _lost_.19:17
andreas hello how to save and quit?19:17
thiago_home please use an editor you know how to use19:17
schierbeck gitte: thanks! i think it19:18
i think it's okay for now to just delete the dir19:18
since it's a plugin, the user won't likely have altered it19:18
gitte andreas: I just thought if you read any tutorial, and that missed out on describing how/when to push, then I'd like the author of that tutorial to fix it. I mean, it is real basic.19:19
← Jacolyte left19:20
schierbeck gitte: hmm, i still need a way to remove the submodule from .gitmodules19:20
gitte schierbeck: ah, right. You'll also need to remove it from .git/config.19:21
schierbeck damn19:21
gitte schierbeck: easy.19:21
← dreiss left19:21
gitte schierbeck: git config --remove-section submodule.<name>19:21
→ dreiss joined19:22
schierbeck gitte: nice!19:22
gitte schierbeck: git config -f .gitmodules --remove-section submodule.<name>19:22
schierbeck that's awesome, thanks!19:22
gitte schierbeck: np19:22
andreas i checked the guides and didnt find anything for this19:23
I'm stuck gher19:23
→ Ratler joined19:23
bremner andreas: you should probably choose a different editor. :wq should save and quit, but life is too short to learn vi on IRC19:24
gitte andreas: you're right, the first Google hit, gittutorial.txt, does not even mention git push.19:24
andreas gitte realy?19:25
gitte Anybody feels like sending a patch?19:25
j_engelh anybody feel like creating the patch in the first place?19:25
andreas I just want to know two or three things come on dont be that lazy to mne19:25
charon gitte: to google? ;)19:25
gitte charon: *giggle*19:26
At least the last sentence of the section says something about "push": http://www.kernel.org/pub/software/scm/git/docs/gittutorial.html#_using_git_for_collaboration19:26
→ ceej joined19:26
← andreas left19:27
← ashleyw left19:27
← mcella left19:29
BeerSerc make install still failing in the templates folder here: http://rafb.net/p/F7SveO31.html19:30
→ markelik_ joined19:30
← ChugChug left19:30
BeerSerc any ideas anybody? I think theres something wrong with the bash/tar oneliner there19:31
→ twas joined19:31
BeerSerc using GNU bash, version 3.2.33, tar (GNU tar) 1.2019:31
→ bieneff joined19:32
charon try just '(cd blt && tar cf - .) >/dev/null' and see what, if anything, it complains about?19:32
j_engelh BeerSerc: goddammit. *USE* option dashes.19:33
-xfo not xfo19:33
and at best also -cf not cf19:33
jast that's not his commands, it's what the git makefile does19:33
j_engelh headdesks19:33
← twas left19:33
gitte thinks that j_engelh makes a problem where there is none; tar xf works quite nicely.19:34
BeerSerc j_engelh: grep -2 "cd blt" templates/Makefile19:34
(cd blt && $(TAR) cf - . | \ cd '$(DESTDIR_SQ)$(template_instdir_SQ)' && umask 022 && $(TAR) xfo -)19:34
j_engelh and just what option is -2?19:34
Run Oops... how can undo a 'git add'?19:34
BeerSerc j_engelh: man grep19:34
gitte BeerSerc: why don't you try "(cd blt && tar cf - .)" yourself?19:35
j_engelh BeerSerc: it's not mentioned.19:35
gitte Run: git rm --cached <file>19:35
jast Run, to get the file in the index back to its state in HEAD, you can use git reset HEAD -- file19:35
what gitte just suggested removes it completely from the index19:35
take your pick19:35
j_engelh meh -2 is actually -B219:35
stupid flags19:35
er, -A2 -B219:36
Run jast: it was an untracked file, I want to keep it.19:36
gitte jast, Run: yes, I assumed that the "git add" was done on a previously untracked file...19:36
→ LuboC joined19:36
jast gitte, yes, i could guess :)19:36
in that case, go with rm --cached19:36
Run ok19:36
→ Weems joined19:36
← Weems left19:36
gitte BeerSerc: So what does that tar cf - . create, that tar xvf - does not think it is a tar archive?19:36
jast who knows, it's from the Makefile in templates/19:37
BeerSerc gitte: one second please, will try19:37
gitte BeerSerc: If that still looks like a tar archive, maybe your tar prefers xof instead of xfo?19:38
thiago_home xfo is ok19:38
-xfo is not19:38
j_engelh so use -xof19:38
that old scheme is so damn confusing it should have been eradicted long ago19:39
thiago_home -xfo means "read file o"19:39
gitte thiago_home: it could also be that that tar does not even understand the -o option.19:39
← coderdad left19:40
BeerSerc (cd blt && tar cf - .) does not complain at all19:40
gitte And it does not output anything, either?19:40
BeerSerc it does output a lot of stuff, I will just check if that is a valid tarball19:40
if I use a filename instead of -, and tar -xfv it afterwards, it works19:42
jast try xof instead of xfo19:42
BeerSerc same error19:42
jast what tar are you using?19:43
BeerSerc tar (GNU tar) 1.2019:43
jast that's weird19:44
gitte What if you insert "tee /tmp/tst.tar |" in the command?19:44
BeerSerc where exactly? you mean tar cf /tmp/tst.tar | tee /tmp/tst.tar | tar xof -?19:45
jast yes19:45
well, no19:45
gitte Actually, I meant after "(cd blt && tar cf - .) |"19:45
bremner BeerSerc: is /path/to/git/git/templates sanitized?19:45
BeerSerc what exactly do you mean by sanitized? I exchanged my local home folder name with /path/to if thats what you mean19:46
gitte BeerSerc: The idea being to find out write out the stdout of the tar cvf - . so that we will see what the tar xfo gets as stdin19:46
bremner Yeah, that's what I meant. Ok. no comment then. /usr/local/share/git-core/templates gets created?19:47
BeerSerc hm, it has been created by a version before, so it is definitely there19:47
← Weasel[DK] left19:48
→ priidu_ joined19:48
← priidu left19:49
← priidu_ left19:49
→ priidu joined19:49
BeerSerc gitte: the tee thing doesnt work, but it should be kind of the same like using a filename instead of -, right19:49
got to leave for 15 min, later19:50
gitte BeerSerc: the idea is not that it will work right now, but that we can see what /tmp/tst.tar looks like.19:50
→ tango_ joined19:51
→ dmiles_afk joined19:54
Run What does this mean:19:56
hikaru:master:slviewer>git reset HEAD -- debian/tg2quilt.mk19:56
debian/tg2quilt.mk: needs update19:56
gitte Means you have uncommitted changes.19:56
In debian/tg2quilt.mk19:57
Run I want to undo those changes to this file :/19:57
gitte You probably wanted to clean them.19:57
Right.19:57
git checkout HEAD -- debian/tg2quilt.mk19:57
Run Ah :)19:57
I'm constantly making changes while on the wrong branch :(19:58
bremner Run: OT, but did you see the topgit 0.5 announcement? topgit now ships with a makefile snippet19:58
Run Then it takes me forever to clean up the mess19:58
bremner (OT, because it ships in debian)19:58
→ ashleyw joined19:58
bremner Run: could git stash help?19:58
gitte Run: you can switch branches with uncommitted changes.19:59
Run I'm not sure what git stash does with untracked files :/19:59
gitte Run: it does not touch them.19:59
Run I have a branch 'topic/pandora/add_support_for_libcwd' that contains new files, committed and all (locally), and now I am working in the branch 'build', which has MANY patches on top of master (the base of the first branch). It didn't build, so I started to fix that, but what I fixed should go into topic/pandora/add_support_for_libcwd... and now I added it to build.20:00
The problem is that I changed files that are untracked in 'build' and committed in topic/pandora/add_support_for_libcwd. It refuses to merge...20:01
it's a mess20:01
← patmaddox left20:02
bremner add lots of things, then stash?20:03
→ kalila joined20:04
bremner Run: by the way, I find git cola (maybe git gui is just as good) handy as it keeps track of the current tracked/changed state of files20:05
kalila hello - How can I list the differences between my working tree (where I committed multiple times) and the tree I "cloned" from?20:05
Ilari kalila: 'git diff origin/master'?20:05
kalila thanks...20:06
Ilari kalila: Or actually more like 'git diff `git merge-base HEAD origin/master`'...20:06
kalila And how to commit a whole branch with its name to the cloned tree?20:08
← Voker57 left20:10
← Ratler left20:10
→ Ratler joined20:10
← komplex left20:11
bremner kalila: do you maybe want a tag? man git-tag20:12
Gitbot kalila: the git-tag manpage can be found at http://git.or.cz/man/git-tag20:12
kalila thanks!20:12
← jackdempsey left20:13
← hurikhan|Work left20:13
→ jnareb joined20:13
jnareb I wonder what was the reason behind allowing to tag (as in: tag object) any type of object, be it commit, blob, tree or another tag...20:15
jast probably "because we can" :)20:15
→ advany joined20:16
jnareb well, tag to blob is used for things like junio-gpg-key but I'm not sure if it is not finding porpose for this ability...20:17
vmiklos it's more about "git is a filesystem as well", imho20:19
jnareb perhaps gittus or gitster could reply to that... pity that they aren't here...20:20
kalila I don't understand what the manuals mean by " index"20:20
vmiklos gittus?20:20
Run Hmm, it's Saturday right?20:20
.. two days no food :/20:20
jnareb vmiklos: if I remember correctly this is IRC nick (handle) of Linus on this channel...20:21
vmiklos that's not too healthy20:21
jnareb: ah. never seen him here :) (probably i'm too young)20:21
Run I'll eat bread, but the shops are closed - as usual...20:21
BeerSerc gitte: http://omploader.org/veTh320:22
vmiklos 04:59 < spearce> linus is almost never on irc. when he is, he's usually called gittus.20:22
uhum :)20:22
BeerSerc this is what comes out, and I cannot untar it20:22
spearce damn, that's from way back in time.20:23
BeerSerc I used (cd blt && tar cf - .) > /tmp/tst.tar to get this20:23
gitte It's a sign of good and flexible design that the data structure does not limit anything when it is not necessary.20:23
← iulian left20:23
→ Voker57 joined20:23
jnareb gitte: true20:24
← ceej left20:24
gitte BeerSerc: aarrgh!20:24
BeerSerc: never, ever export that variable!20:24
BeerSerc which variable?20:24
jnareb I'm asking that because curently GitHub doesn't support such non-commit tags well (they are displayed correctly, but canot follow to destination, IIRC).20:24
CDPATH ?20:25
gitte BeerSerc: CDPATH.20:25
BeerSerc oh... I see where the problem could be ...20:25
;)20:25
gitte BeerSerc: the first line of tst.tar is the directory path.20:25
jnareb http://git.or.cz/gitwiki/GitFaq#head-9972ad5add19d3fd22c803dcdddb39ce6dd3b6af20:25
BeerSerc gitte: right20:25
jnareb Gitbot: faq CDPATH20:25
Gitbot jnareb: FAQ entry 'CDPATH' not found.20:25
← Bass10 left20:26
kalila Is there a command that commits (to local tree) and pushes (to remote treee) at the same time?20:26
jnareb kalila: why would you want that?20:26
jast no. that's always two commands. feel free to define an alias if you must have it.20:26
jnareb kalila: the separation between commit and publish in distriobuted VCS is there for a reason...20:27
kalila, jast: or you can use post-commit hook...20:27
jast ew.20:27
→ Ratler_ joined20:27
kalila I do understand that, it's just that I'd like to do it often20:28
fine20:28
gitte BeerSerc: did nothing in the Makefile complain about CDPATH being exported?20:29
BeerSerc gitte: nope20:29
→ Ace_NoOne joined20:29
Ace_NoOne hi - is there a simple way to check out a Hg repo with Git? doesn't need to be as comprehensive as git-svn20:30
gitte BeerSerc: Ah right, I thought that the Makefile checked for that... it does not.20:30
BeerSerc maybe it should20:31
bremner Ace_NoOne: tailor?20:31
BeerSerc I am not using it for so long...20:31
jnareb Ace_NoOne: see tools mentioned in http://git.or.cz/gitwiki/InterfacesFrontendsAndTools20:31
Ace_NoOne thanks20:31
jnareb Ace_NoOne: or best the one mentioned in http://vcscompare.blogspot.com20:31
Ace_NoOne hmm20:32
that's a lot of posts to dig through there20:33
jnareb I meant here the post (at vcscompare) titled "Import tools" (one of very first posts)20:34
Ace_NoOne ahh thanks, that'll do20:34
← ph^ left20:35
Voker57 the main idea of http://vcscompare.blogspot.com appears to be "git rocks, bzr sucks" :D20:37
← Ratler left20:37
bremner fairandbalancedvcs.blogspot.com is still available I bet :-)20:38
jnareb Voker57: actually I think it was scientific comparison and benchmarking (as far as it is possible given the differences between DVCS)20:38
→ coderdad joined20:38
jnareb Voker57: pity that Pieter didn't have time and energy to do the rest of benchamrks (admittedly the more difficult ones): speed / performance, and network (latencu, throughput, total bytes transferred, protocol over20:40
Run Is it possible to work locally in two different windows on two different branches, but share (local) commits somehow?20:42
→ swhitt joined20:42
Run Thus, if I commit in one window on branch master, and I go 'git merge master' in the other, then I'd get that commit ?20:42
s/go/do/20:42
→ carllerche joined20:43
← tvw left20:43
jast yes, but it's not too enjoyable20:45
you can set an environment variable that tells git what dir to use at working tree and what dir to use as the git dir (see man git)20:45
so you could set two different working tree variables20:45
hmm, and you'd have to use different index files too20:46
and different HEAD files...20:46
best to just clone the repo, i guess.20:46
Run But then I wouldn't get things unless I push20:46
jast git will reuse the pack files if possible20:46
well, does it matter if you use git merge or git pull?20:46
git pull otherrepo foo. done.20:47
Run Oh, no push thus20:47
jast yeah20:47
← Sigma left20:47
jast you shouldn't push to a non-bare repo anyway (see faq non-bare)20:47
Gitbot jast: Pushing to non-bare repositories is discouraged. See http://git.or.cz/gitwiki/GitFaq#non-bare20:47
← yann left20:50
→ scientes joined20:51
→ Ratler joined20:52
← Ratler_ left20:52
← Voker57 left20:54
→ devurandom joined20:54
devurandom Hey!20:54
j_engelh mvs devurandom back to /dev20:55
devurandom :)20:55
← apeiros_ left20:55
→ apeiros_ joined20:55
devurandom Is it possible to have 2 branches in a repository which did not start at a shared common point in the past, but will later be merged eventually?20:55
jnareb Run: contrib/worktree20:56
← kalila left20:56
← coderdad left20:56
gitte devurandom: yes.20:56
jast yes, see "create empty branches" on book.git-scm.com20:56
jnareb devurandom: yes, it is, although I'm not sure if you won't have to do first merge using low-level tools anyway...20:56
→ jm joined20:56
← dhutty left20:56
gitte jast, jnareb: no trickery is needed.20:57
jnareb devurandom: for example git repository has quite a bit of roots (parentless commits): git core, git mailing tools, gitk, gitweb, git-gui20:57
← bieneff left20:57
jast yeah, you can avoid it by merge two separate repositories20:57
gitte AFAIK only gitk was merged via the "coolest merge ever"20:57
jast that sounds interesting20:58
jnareb aren't both gitk and git-gui merged using 'subtree' merge strategy?20:58
moccuo how come when i delete or rename certain files, they are still there when i do a commit20:58
tango_ is it possible to use git svn to track a signle FILE in a svn repo?20:58
jast no.21:00
unless it's the only file in a directory21:00
tango_ k21:01
devurandom Hm, maybe I'll leave it like this then, and just merge the repositories later...21:01
moccuo how do you remove/rename files and have the old ones deleted in the commit?21:01
→ flaguy joined21:02
jnareb moccuo: git-mv or git-rm ?21:03
moccuo: but there are a few rare situations where when switching branches old deleted files would persist because git is hesitant to remove things which _might_ be unrecoverable21:03
moccuo oh kk21:04
hmm, its just i delete stuff from within textmate quiet often. i guess ill just do it in the terminal from now on, heh21:04
or rename stuff21:04
→ coderdad joined21:05
jnareb moccuo: removed files should be picked by git automatically if you do "git commit -a"21:05
moccuo aah, alright21:05
jnareb moccuo: and there is "git add -u" and "git add -A"; see man git-add21:06
Gitbot moccuo: the git-add manpage can be found at http://git.or.cz/man/git-add21:06
← tritz left21:06
moccuo i always just did a "git commit"21:06
← x_or left21:06
→ Ratler_ joined21:06
moccuo alright, thanks jnareb21:06
jnareb "git commit" picks from index. Line "rm file" removes only from working directory. So git-rm / git-rm --cached is needed21:06
← Ratler_ left21:09
moccuo but if i do "git commit -a" from now on i dont really need git-rm ?21:11
→ careo joined21:11
→ x_or joined21:11
← apeiros_ left21:12
← ajonat left21:13
tango_ jnareb, hey there. I'm waiting for a reply from juno for the allheads stuff, about the 'grab all the refs at once' and the rate limiting. of course, if you have comments too, they'd be appreciated21:14
← Ratler left21:17
→ davi joined21:17
← coderdad left21:22
→ Nebukadneza joined21:30
Nebukadneza heyho21:30
j_engelh all heads battle station - this is a drill this is a drill!21:31
gitte Hi Nebukadneza21:31
Nebukadneza when i have cloned a kernel git repository, compile something in it and then do a git checkout <some other revision>, do something in the other revision and then switch back to the first revision again ... would my compiled .o's still be there?21:32
gitte Yes.21:32
They are untracked, and therefore untouched.21:32
Nebukadneza cool21:32
albertito Nebukadneza: yes, git will NOT touch any files that you've not explicitly added to the repository21:32
Nebukadneza :)21:32
i mean21:32
are they there in the "middle checkout" too?21:33
albertito Nebukadneza: yes21:33
Nebukadneza duh21:33
ok21:33
:)21:33
← ankit9 left21:33
Nebukadneza so to have them tied to one checkout i'd have to add it and check it in?21:34
hell21:34
→ riddochc joined21:34
Nebukadneza i think i need to understand tags and branches in git first ;)21:34
← Beket left21:36
→ bieneff joined21:37
→ sanginus` joined21:37
← sanginus` left21:38
albertito Nebukadneza: yes, but it's really strange that you will have such a need. I think it's more common to only check in "source files" (whatever they may be) and leave content that is automatically generated out of the tree21:38
→ Ratler joined21:38
Ilari Nebukadneza: Changes in working tree and index are carried over when switching branches (assuming those changes don't prevent switch). There's also (destructive) mode that doesn't carry over the changes.21:38
→ DooMSonG joined21:40
→ doener joined21:40
Ilari Nebukadneza: Actually, that distinction is only for files that exist in both branches...21:40
→ john-cena joined21:40
john-cena PEDALE TUK NA KAV EZIK GOVORITE VASHTA MAMICA DEBA21:40
jast why don't spammers ever speak the same language as their targets?21:41
john-cena ||-||doomsong||-|| :21:41
DooMSonG jast hello gay:)21:41
Nebukadneza hm21:41
lemme try21:41
DooMSonG Êàæè âå ïåäåðàç21:41
:Ä21:41
j_engelh slaps DooMSonG with utf-821:41
DooMSonG ?21:41
jast DooMSonG, how did you know?21:41
DooMSonG ê'âî èñêàø ìàöå21:41
j_engelh you're just printing gibberish21:42
DooMSonG Êî ðå÷è?21:42
riddochc A thought I had: If I put my tests in a separate git repository (possibly linked as a subproject) would it make it easier to automate git bisect runs for failing tests?21:42
DooMSonG oooo21:42
john-cena MAMAVI DEBA CIGANSKA DEBA OGIN DA VI IZGORI DEBA VASHTA MAMA DEBA21:42
ChanServ set mode: +o21:42
DooMSonG ok21:42
i speak english21:42
john-cena ops21:42
i speek english21:42
YeSsS21:42
j_engelh -.-21:43
DooMSonG :)21:43
i'm from bulgaria21:43
;]21:43
gitte john-cena, DooMSonG: anything sensible you have to contribute?21:43
cehteh huch Nebukadneza you here? :)21:43
john-cena YeSsS baby21:43
DooMSonG We're from bulgaria:))21:43
john-cena YeSsS baby21:43
:D21:43
j_engelh apparently not.21:43
Nebukadneza cehteh: sure i am21:43
riddochc My question is, of course, for anyone interested in actually talking about git...21:43
john-cena am bulgerian21:43
Nebukadneza cehteh: i'm slooowly starting to move towards git21:43
DooMSonG We are Bulgarian21:43
john-cena im bulgerian21:43
cehteh finally :P21:44
DooMSonG Bulgarian21:44
;]21:44
john-cena my names is DANIEL21:44
Nebukadneza cehteh: now i have a NEED for git21:44
j_engelh uh this is not your aol chat.21:44
DooMSonG my name is Dimitar21:44
and i love fight21:44
:P21:44
Nebukadneza before i didnt need it21:44
gitte kicked DooMSonG (User terminated!)21:44
→ DooMSonG joined21:44
drizzd riddochc: well, git.git has its tests in git.git. And I don't see why submodules would help that21:44
gitte kick john-cena21:44
gitte kicked john-cena (User terminated!)21:44
→ john-cena joined21:44
DooMSonG Low!21:44
cehteh DooMSonG: john-cena ... you know how /ignore works?21:44
DooMSonG shit21:44
gitte kicked DooMSonG (User terminated!)21:44
john-cena what21:44
→ DooMSonG joined21:44
gitte set mode: +b21:44
gitte kicked john-cena (User terminated!)21:44
→ john-cena joined21:44
gitte set mode: +b21:44
gitte kicked john-cena (User terminated!)21:45
jast hmm, they're not registered, so they can change nicknames if they're serious about annoying people21:45
let's hope for the best21:45
cehteh heh21:45
git support closed for today ... at least for some :)21:45
j_engelh they were not asking about git anyway21:45
gitte Anybody knows how I could ban people who have a certain numeric IP in their user name?21:46
ciaranm they were being gits21:46
gitte: *!*@1.2.3.421:46
gitte ta21:46
Hey, if somebody tells me they love to fight, I have a reputation to lose if I do not react.21:46
→ masak joined21:46
j_engelh most clients have /kb which suffices21:47
cehteh just suggests /ignore ..21:47
masak I have a remote branch 'dispatch' that I want to download to a branch 'dispatch' in my local repository. what command do I use?21:47
cehteh or my anti troll plugin .. but thats another thing21:47
jnareb tango_: about using one versus two calls to git-for-each-ref when enabling 'remotes' to be shown?21:47
tango_ jnareb, yep21:48
in summary view21:48
uau masak: just fetch the current state once? or do you expect to track the same remote branch later?21:48
jnareb tango_: I don't think that _one_ more call is important here21:48
tango_ it's radically different because of the 16 refs limit21:48
riddochc drizzd: This is to deal with the situation where the test isn't present in the particular version bisect is trying to check. Or is there a better way?21:48
tango_ junio seemed to think it was 8-D21:48
jnareb tango_: that is one call per commit when displaying 100 of them per page that you have to care about ;-)21:48
masak uau: I think I expect to track it later. I might also do commits on it myself.21:48
tango_ jnareb, ok21:48
masak uau: would `git fetch origin dispatch` cut it?21:49
jnareb tango_: if somebody decides that it matter, he/she can always change it "in tree"21:49
riddochc drizzd: an alternative would be to make another working directory to run the tests from...21:49
drizzd riddochc: you can always 'git checkout master test'21:49
tango_ ok21:49
uau do you already have that remote repository as a remote named "origin"?21:49
masak uau: yes.21:50
← DooMSonG left21:50
uau then you should already have it as a remote branch21:50
origin/branchname21:50
← jnareb left21:50
masak uau: I might. it doesn't show up in `git branch`, though21:50
then again, neither does any of the origin/foo branches21:50
uau without arguments "git branch" only shows local branches21:50
masak aha.21:51
uau you need to use -r or -a to see remote ones21:51
masak ok, I see it.21:51
now what do I do?21:51
uau depends on what you want to achieve21:51
if you want to check out the contents/history then you can do that with the remote branch21:51
if you want to do development based on it then you probably want to create a local branch based on the current state of the remote21:52
masak uau: and if I want to make my own commits...21:52
riddochc drizzd: but doesn't bisect need to change the state of the working directories in order to run the test on older versions? Making another branch won't keep the newer tests in the working directory while also testing older code in the same directory. That's the issue.21:52
masak uau: as in `git co -b origin/dispatch dispatch`?21:52
uau yes21:52
masak thank you.21:52
uau actually wrong order of arguments21:53
git checkout -b dispatch origin/dispatch21:53
masak right.21:53
...this is why I asked, I do this too seldom :)21:53
Ilari And in late enough versions (IIRC, newer than anything released), 'git checkout --track origin/dispatch'. :-)21:53
masak heh :)21:53
now let's say this branch was branched from master, and I want to list all the commits since then. how do I do that? `git log ...`?21:54
Ilari Also, some old versions require '--track' for tracking info to be written into configuration.21:55
masak: 'git log master..dispatch'?21:55
gitte Ilari: since v1.6.0-rc2-30-gbb0ceb6 it is supported.21:55
masak Ilari: sounds like what I want, yes.21:55
what's the difference between master..dispatch and master...dispatch in this case?21:56
uau masak: it shows commits that exist in the second argument but not in the first one21:56
Ilari masak: 'master...dispatch' also covers commits in 'master' that occured after the fork...21:56
uau oops misread it somehow21:56
masak oki21:57
→ sanginus` joined21:57
Ilari gitte: That goes from tag into future... What does --contains (go backwards) give?21:57
← exon left21:58
pflanze What should I do with filter-branch to remove empty commits from merges? this won't do that: if [ z"$1" = z"$(git rev-parse "$3"^{tree})" ] then skip_commit "$@" ..21:58
Merge commits can still have empty commits as ancestors.21:58
Ilari fatal: cannot describe 'bb0ceb6264fa1aea6e68e07cb13cd9a88473febb'...21:58
Gitbot [git bb0ceb626]: http://tinyurl.com/5adhyd -- checkout --track: make up a sensible branch name if '-b' was omitted21:59
pflanze empty commit==commits that don't change anything I mean. hum. Or really empty actually, I guess that's the reason, that's something else.21:59
thiago_home pflanze: git-commit refuses to create empty commits21:59
so how were they created in the first place?21:59
pflanze thiago_home: by filter-branch, using git rm --cached.22:00
thiago_home if you removed files, the commits aren't empty22:00
pflanze in this branch they are22:00
because this is a merge of an independent branch (no common ancestor),22:00
and I rm everything from the beginning in that branch.22:01
masak is there any way to show the logs in increasing date order instead of decreasing?22:01
pflanze so the commits are all equal and all empty. But still merged.22:01
thiago_home masak: --reverse22:01
masak ah, just found it :)22:01
thiago_home: thanks22:01
→ intripoon joined22:01
pflanze I have to change or rather drop the merge commit now.22:01
actually, not drop it, that merge commit contains new non-removed files. strange, going to investigate22:02
→ mitchell joined22:03
masak combining logs and diffs...?22:03
← mitchell left22:04
Ilari masak: 'git log' has '-p' option to show diffs...22:04
masak Ilari: thank you.22:04
I just want to note that you guys give awesome service.22:04
pflanze ps. regarding emptyness: I'm talking about the commits having empty 'tree' objects. With sha1 4b825dc642cb6eb9a060e54bf8d69288fbee4904 which google readily confirms that it's the empty tree :)22:05
Gitbot [git 4b825dc64]: http://tinyurl.com/55rga7 [tree]22:05
Ilari pflanze: Can empty commit have non-empty ancestor?22:08
← LiamH left22:08
robinr sure22:08
pflanze Ilari: you mean in general? Why not.22:08
Ilari Meant in that practicular case...22:08
← naeu left22:09
thiago_home pflanze: that means it's a commit where all files were removed.22:09
pflanze No. This branch has everything removed22:09
Fun thing is:22:09
thiago_home pflanze: why do you want to remove that information>22:09
Ilari pflanze: Then you could use grafts to modify history and then make it permanent using filter-branch...22:09
pflanze that branch was merged in a commit which *is* dropped in my new history: but the next newer commit now has the *first* of all commits of that other branch as additional parent.22:10
So, I've got a commit which had 1 parent previously but now has 2 parents, one of them the first commit of the useless branch, and being empty.22:10
I don't think graft would help as I need to modify a commit that I want to keep. By reducing the number of parents to one.22:11
← d0k left22:11
pflanze Or could 00000000.. be used as magical "drop that pointer" sha1?22:11
Ilari It is possible to use grafts with filter-branch for performing quite complicated rearrangements of commits (including dropping them)....22:11
pflanze I know I can replace pointers.22:12
→ andreas_ joined22:12
pflanze But not how to remove them, with grafts.22:12
charon pflanze: the grafts file specifies the full parent list of a commit (if it is listed there)22:12
andreas_ TML here?22:12
charon pflanze: so if you drop a parent from the list in the grafts file, it 'disappears'22:12
Ilari pflanze: It is possible to add, remove and replace parents using grafts...22:12
pflanze hm I should re-read that grafts stuff, ok22:13
Anyway, since I'm building a script to help me handle repository splitting, I'd like to find out how do do it right in the first place.22:13
Ilari pflanze: 'git-split'? Anyway, there's --subdirectory-filter...22:14
pflanze So that filter-branch will make it right without me having to fiddle with grafting manually afterwards.22:14
→ geckosenator joined22:14
pflanze checks22:14
geckosenator how do I print the entire contents of a file given the commit and filename?]22:14
thiago_home git show commit:filename22:14
pflanze I'm using perl to decide with regexes on whether a path is being kept.22:15
That's more flexible than subdirectory filter.22:15
← intripoon left22:15
pflanze sighs and googles for git-split22:15
geckosenator thiago_home: I get an error saying to use '--' to separate paths from revsions22:16
→ intripoon joined22:16
← sanginus` left22:17
thiago_home geckosenator: not in this case. It's a colon.22:17
Ilari geckosenator: Likely mistyped commit reference or mistyped path...22:18
thiago_home personally doesn't like that construct because it doesn't easily translate to path completion22:18
thiago_home especially in large repositories22:19
→ Wyk3d joined22:22
← Kangou left22:23
Wyk3d hello .. does anyone know a way to use mergetool for solving the conflicts that occur after applying a patch with git apply ?22:24
← ashleyw left22:30
→ LotR joined22:31
pflanze How does one drop a remote branch pointer from a local repo?22:32
→ ankit9 joined22:32
pflanze git branch -d says not found, and git remote rm will not accept someremote/somebranch as argument22:32
j_engelh git remote prune remotename22:33
pflanze fatal: 'home-Adressen/home': unable to chdir or not a git archive22:33
← andreas_ left22:33
pflanze I only wanted to drop some. Well, maybe I'll just drop the whole remote after merging.22:34
Ilari pflanze: It takes name of remote, not of remote branch.22:34
pflanze I need to merge one branch first.22:34
I just expected I could drop the uninteresting branches like local branches.22:34
but makes sort of sense not being able to. anyway.22:34
← devurandom left22:37
← anttih left22:37
→ parasti joined22:37
→ joshdavey joined22:38
Arachnid If I delete a branch, is any record of it retained? For example, will github's network view still show the commits to it, merges to other branches, etc22:40
?22:40
Ilari Arachnid: Deleting a branch via 'branch -d' only affects the repository it is done in. There's also remote delete (git push <repo> :<branch>). Reflog of HEAD can still reference commits in that branch. Also if it is merged to some other branch, the merged commits don't disappear.22:44
pflanze he, just found a gitk bug. git version 1.6.0.422:45
any gitk developer here?22:45
→ andreas_ joined22:45
← andreas_ left22:45
→ sergiodj joined22:45
Ilari pflanze: How to reproduce, what you expected and what you got could be useful...22:46
pflanze gitk jumps to the wrong location when I click on a Parent link in a commit the first time.22:46
→ andreas_ joined22:46
pflanze It jumps to a line above the correct one.22:46
When I go back and click again, it jumps to the right place.22:46
It's 100% reproducible.22:46
→ mosno joined22:47
andreas_ hey guys i got a lil exception here maybe someone is known to this it came from my kgit QObject::connect: No such slot kgitView::slotGitSendCommitDiff(KProcess*) QObject::connect: (sender name: 'unnamed') QObject::connect: (receiver name: 'kgitViewbase')22:47
pflanze i.e. after each restart of gitk, the first click on that Parent gets carried out wrong.22:47
→ jackdempsey joined22:47
→ botanicus joined22:47
andreas_ Can someone tell me what that means or what that produces?22:47
← ankit9 left22:48
→ binjured joined22:50
← Weaselweb left22:50
← bieneff left22:51
drizzd pflanze: I think linus mentioned a bug like that on the list before22:59
pflanze good22:59
← muthu_ left22:59
drizzd IIRC the issue is that the lines to jump to are computed and then the commits are reordered in some way23:00
← softdrink left23:05
→ ankit9 joined23:06
← advany left23:06
→ seangrove joined23:08
← andreas_ left23:09
→ andreas_ joined23:11
andreas_ how long doesit take to upload a patch tio github?23:12
→ Eridius_ joined23:12
jast usually not very long23:12
unless it contains load of huge files23:12
← seangrove left23:13
← earcar left23:13
← andreas_ left23:15
← masak left23:16
← Ace_NoOne left23:17
→ prophile joined23:21
prophile how does one push to a remote creating a new branch?23:21
git push <remote> <my-branch-name>:<remote-branch-name> right?23:22
gebi push supports the --dry-run flag ;)23:22
← LuboC left23:24
prophile i'm sorry, i don't understand what i'm missing wrt --dry-run23:24
it's just that that's not working for a friend of mine23:25
and I'm not sure why and I was asking if the syntax is right and something's broken on the remote or something23:25
→ bgerber joined23:26
gebi yes the syntax is right23:26
← bgerber left23:37
← riddochc left23:42
← solydzajs left23:43
drizzd prophile: well, the exact syntax would be "git push <remote> <branch>:refs/heads/<branch>23:43
prophile thanks23:43
drizzd but I think only relatively old versions of git don't allow to leave out the refs/heads part23:43
prophile yeah, the former syntax has worked for me but I tend to use fairly up-to-date versions23:44
→ bgerber joined23:46
← EmilMedve left23:48
→ softdrink joined23:51
→ gottesmm joined23:54
→ solydzajs joined23:58

Logs Search ←Prev date Next date→ Channels Documentation