shruggar	ncopa, how about whenever a new tag is pushed, Have A Script check-out that tag, build it, etc? If you only want certain tags to be auto-built, then have a naming convention for tags
Ilari	ncopa: The script could also export a snapshot, unpack it somewhere and build.
ncopa	hum
ncopa	for each git push i already git pull and build the changes
Ilari	ncopa: Using git pull in scripts is a bad idea.
shruggar	ncopa, so do it for tags as well. What's the actual issue you're having?
ncopa	detect when there is new tag
ncopa	lunch. brb
shruggar	as was said before, git-fetch will tell you about new tags it fetched
Ilari	ncopa: Also in post-update hook, updated/new tags are visible in list of refs updated.
Ilari	New way to abuse remote helpers: As client-side "post-push" hooks. ):-)
shruggar	I suppose having a real post-push makes as much sense if not more sense than post-commit
tasslehoff	How can I fetch objects from one local repo to another? Setup a remote? None of the repos are bare.
shruggar	tasslehoff, bare repositories are only a convenience and are not required. Just set up a remote
shruggar	git remote add foo /path/to/repos
tasslehoff	shruggar: thanks
tasslehoff	when I add the remote I get a warning about "no common commits", is that very bad?
shruggar	not necessarily, but it's probably not what you want
shruggar	tasslehoff, how were these repos created?
tasslehoff	shruggar: git svn clone :)
tasslehoff	shruggar: we're moving to git, but I still have to work with the svn server
shruggar	tasslehoff, I'm guessing that the repos weren't created using the same "git svn clone" string. For example, one may have been created with "git svn clone svn+ssh://foo/bar/baz"  while another was created with "git svn clone file:///bar/baz"
tasslehoff	shruggar: yeah. one is actually trunk, while the other one is branches/foo.
tasslehoff	hmm, when I say it out loud, this sounds sketchy
ziro`	what should i set my autocrlf to?
ziro`	i want to only work with LF's but I want to ensure I don't accidently add CRLF's
shruggar	tasslehoff, there's the problem, then :)    svn branches don't translate neatly into git branches like that. In some situations, you can clone the entire svn repository (specifying where branches, tags, etc, go) and then treat them as normal git branches
shruggar	tasslehoff, I've yet to meet a mature svn repository (out of all three I've looked at, so my own experience is rather minimal) where git-svn could process that sort of thing sanely
tasslehoff	shruggar: hm. I also want to add a remote to my colleagues checkout of the same branch, but I sense that will cause the same problem
shruggar	tasslehoff, if you both cloned using the same string, you should be able to pass commits back-and-forth no problem
Ilari	ziro`: input then
shruggar	tasslehoff, for example, I pass commits through a bare repository between work and home, since work and home both cloned via the same git-svn string
bob_f	sitaram: Sorry, juggling a bunch of things at once here. When I try to "git clone git@server:gitolite" I get "fatal: 'gitolite' does not appear to be a git repository".
bob_f	And same if I do "gitolite-admin", which is what the repo is called.
Ilari	bob_f: Try 'gitolite:gitolite-admin'.
Ilari	bob_f: (weird games gitolite plays with ssh).
tPl0ch	Hi, I am trying to create a post-receive hook on our server. It is as simple as that: http://pastebin.com/hW0Ffq0T If I run it as the 'git' user, it works flawlessly, but when pushing via git+ssh from the local repo, it does not get executed.
bob_f	Ilari: Nope, same error.
Ilari	tPl0ch: 1) Don't run git as root. 2) Don't use pull in scripts, 3) if you cd in script, remember to unset $GIT_DIR.
bob_f	This is probably in the docs, let me check it up.
Ilari	bob_f: 'git clone gitolite:gitolite-admin' (literally that way) says that it does not appear to be git repository?
bob_f	Oh, wait.
Ilari	bob_f: As opposed to things like access denied, or host not found.
bob_f	Nope, that dosen't work either - yeah, says it's not a repo.
Ilari	bob_f: What does 'ssh gitolite' (yes, 'gitolite' literally) say?
bob_f	Ilari: It says hello and lists repos and perms.
tasslehoff	and the final straw. it's not possible to use git-svn to create a bare repo, with support for pushing changes back up?
tPl0ch	Ilari, so, how do I update a clone in the post-receive hook?
tPl0ch	without pull?
Ilari	bob_f: 'git ls-remote gitolite:gitolite-admin' (again, literally).
Ilari	tPl0ch: fetch + reset.
shruggar	tasslehoff, "not possible" is too strong a word.   "very inconvenient and annoying to set-up, and always requiring awareness of the set-up" is more accurate
tasslehoff	shruggar: :)
bob_f	£ git ls-remote gitolite:gitolite-admin
bob_f	fatal: 'gitolite-admin' does not appear to be a git repository
bob_f	(gitolite-admin is listed when I do "ssh gitolite" though).
bob_f	And I have R W perms.
Ilari	bob_f: Some very broken gitolite setup? If you didn't have permissions, it would print access DENIED message.
bob_f	Hrm.
Ilari	bob_f: Try 'git ls-remote gitolite:does-not-exist'?
bob_f	£ git ls-remote gitolite:does-not-exist
bob_f	fatal: 'does-not-exist' does not appear to be a git repository
bob_f	fatal: The remote end hung up unexpectedly
Ilari	bob_f: Eeh... For me the error is: "fatal: remote error: R access for does-not-exist DENIED to Ilari".
bob_f	Hmmmm.
Ilari	bob_f: Wonder what the heck is going on. Based on 'ssh', there's gitolite on other end, but git connections look like they bypass gitolite.
bob_f	Yeah, weird, huh ?
shruggar	tasslehoff, the most annoying part of it: commits will ALWAYS be re-written when you dcommit, so if you "git push" to such a repository, a "git pull" right afterwards will try to merge. "git pull --rebase" /might/ work, but something like "git push && git remote update && git reset --hard origin/master" is more what is required
shruggar	tasslehoff, (something like that, anyway)
Ilari	bob_f: (ignore the "fatal: remote error " part, it is artefact of local setup).
bob_f	Ah okay.
Ilari	bob_f: Because error messages from gitolite are converted into protocol errors, and client prints prints protocol errors with 'fatal: remote error: " prefixed...
sitaram	bob_f: putty?
bob_f	sitaram: Nope.
Ilari	bob_f: "ssh gitolite git-upload-pack does-not-exist". What does that complain about?
bob_f	Just £ ssh -V
bob_f	OpenSSH_5.1p1 Debian-5, OpenSSL 0.9.8g 19 Oct 2007
sitaram	bob_f: client OS and sever OS?  (shooting in the dark here...)
tasslehoff	shruggar: hm. plan 1 is now to speed up the process towards !svn
bob_f	ubuntu-server and debian on my end.
bob_f	£ ssh gitolite git-upload-pack does-not-exist
bob_f	fatal: 'does-not-exist' does not appear to be a git repository
shruggar	tasslehoff, it is a good plan.    Note that a new interface to svn is being actively worked on at the moment. I don't know if it will solve any of these problems, but I hardly see the point if it doesn't :)
Ilari	bob_f: So that looks like it bypasses gitolite (but 'ssh gitolite' doesn't...). Hmm...
Ilari	bob_f: You ran 'ssh gitolite' and 'ssh gitolite git-upload-pack does-not-exist' from the same shell (just to rule out environment fscking things up?)
bob_f	Here's my ~/.ssh/config just in case: http://bpaste.net/show/7307/
bob_f	Ilari: Yes, same shell, no env changes.
tasslehoff	shruggar: is that interface mature enough for testing, or is it way-pre-alpha?
Ilari	bob_f: Looks sane...
shruggar	tasslehoff, I think last I saw there were certain aspects of it for which testing was requested, but I wouldn't call it mature enough that it's worth testing for actual use yet
rajeshsr	How does git distinguish between tag name and a branch name, when you do git branch newName tagName etc..
rajeshsr	?
sitaram	bob_f: run src/sshkeys-lint in your gitolite clone (prints USAGE message), then folow instructions, and report back
Ilari	bob_f: Bit of brute-forceish, but: Do 'ssh -vvvv gitolite' and save the full output to file. Then do the same (saving to other file) for 'ssh -vvvv gitolite git-upload-pack does-not-exist' and then diff the two files.
Action: FauxFaux has managed to lose a commit from reflog. /o\
bob_f	Ilari: Sure.
Ilari	bob_f: 'diff -u file1 file2' to diff them.
engla	git diff --no-index file1 file2   and you can use a more familiar diff
bob_f	Ilari: I can pastebin this safely ?
Ilari	bob_f: It shouldn't have any private information.
bob_f	Yeah, looks safe.
bob_f	http://bpaste.net/show/7309/
ncopa	back
ncopa	thanks all
Ilari	bob_f: What the heck? Looks like 'ssh gitolite' did actual login?
Ilari	bob_f: Do the sshkeys-lint stuff...
level09	when using reflog to revert, do I use checkout ? or reset ? or revert ?
wereHamster	what do you mean by 'revert'?
level09	like undo changes
level09	i have merged some branches and did some cleaning that messed up everything
Ilari	level09: 'reset --merge' probably then...
Ilari	level09: Or --hard (dangerous).
wereHamster	was the merge successful?
wereHamster	or did it fail because of conflicts?
level09	failed
wereHamster	reset --hard
wereHamster	or checkout --merge -- .
level09	but if I want to go back into a specific point in my reflog
wereHamster	reset --hard or checkout
level09	okay thanks ! I'll try that
bob_f	Ilari: I'm gonna run to lunch, but this lint stuff should hopefully help me figure it out - thanks a lot for your help though, it's really appreciated.
bob_f	I'll let you know how it goes.
sondreb	when i type "git rm <filename>" it deletes the file but then sends the message "Deletion of directory 'ubs/common/composer' failed. Should I try again? (y/n)?"
sondreb	why does it try to delete te folder?
mathepic	Whats the filename?
sondreb	C:\BuildSys\work\ubs\common\composer>git rm comp_resbuild.pm
sondreb	rm 'ubs/common/composer/comp_resbuild.pm'
sondreb	Deletion of directory 'ubs/common/composer' failed. Should I try again? (y/n)?
jupko	http://pastie.org/1014987
jupko	what's the problem with my local repository ?
jupko	How can I correct it ? You can see the situation if you open the link I pasted.
mathepic	sondreb: Thats odd.
wereHamster	jupko: git fsck --full
jupko	okay
wereHamster	jupko: and before you do that, can you check if the file .git/objects/e1/17cf9010e5973e9d65139424bde2e27cf7389c exists and if it is zero bytes long?
jupko	oops
jupko	I have run fsck before the checking
jupko	4KB
jupko	and it exists
jupko	http://pastie.org/1014995
jupko	you can see the output of fsck operation and the size of that object file.
jupko	the question is still existing, how can I fix it ? :)
jupko	and what can cause such a problem at all ? This will be my other question.
wereHamster	do you have a backup of that repo? or a clone that contains that object?
mathepic	You need to replace that object with one from a noncorrupt repo.
jupko	I can do a new clone because I have pushed the commits
jupko	but I would like to avoid it next time so I would like to know what can cause such a problem.
wereHamster	filesystem corruption for example
jupko	only in this small git repository and no corruption in another places ? :)
jast	no corruption you have found so far
wereHamster	a block on the disk is usually 512 bytes, well within the 4kb file that is corrupt
jast	and it's actually quite likely to happen that way if, for example, you're working on that repository and you get a power failure in the middle of something
jupko	4KB is the page size on i686
wereHamster	page size, yes. But not block on the harddrive
jupko	ah okay I see
jupko	I may know the problem.
jupko	I am developing a driver
jupko	and syscrash occured because of some bad lines.
jupko	that I inserted
jupko	and maybe it caused that.
jupko	meanwhile I was in this repository.
jupko	so it's definitely worth to make a backup directory then.
jupko	:-)
jupko	wereHamster: thank you.
jast	in theory git syncs stuff before updating a branch, but perhaps your filesystem did something nasty and ignored it
jupko	jast: no problem with filesystem at all
jast	well, the corruption has to have been caused by *something* :)
anebi1	hi. i'm newbie in git and i want to ask if is possible to clone a repository under folder located in my current git repository. for example if my git repository is in this http path /git/php, i want to clone a repository under /git/php/vendor. is this possible somehow?
jast	are you, perchance, talking about submodules?
anebi1	jast: if you call this submodule, then it is. i'm looking it from http url side. vendor is a subfolder in my gitrepo and i want to clone all content of remote under this folder.
anebi1	jast: i'm reading now for submodules :)
jast	oh, you want to clone only part of a repository?
jast	if so... well, tough luck. git doesn't support that.
anebi1	jast: the people moved the project from svn to git, i think i will need the full cloning if their repo. i have not checked yet how they have done this.
anebi1	but the content should go under vendor subfolder in my repo
jast	sounds like a case for submodules all right... or the third-party git subtree tool, if you prefer that
Mpenz	sitaram:  :)  Thanks.  I figured.
jast	some people don't like submodules very much, e.g. because you have to manually update the submodule and record the update whenever you need a newer version
anebi1	jast: i see. i will learn how submodules work and probably will go that way. thanks
tPl0ch	Hi, I am still trying to get a post-receive hook to run: http://pastebin.com/Bkj9ULm1. If I run that as the user 'git' on the server, it works flawlessly, but when I push from my local repository, the hook won't execute.
Last message repeated 1 time(s).
PerlJam	tPl0ch: how did you run it from the command line?
PerlJam	tPl0ch: you do have execute permission set, yes?
tPl0ch	PerlJam, yes, it's executable, and I did a su git && /path/to/hook/post-receive
tPl0ch	it did everything right
tPl0ch	but whatever I do, it won't execute when pushing from my local repo
PerlJam	Are you sure you're pushing to the right place?
tPl0ch	PerlJam, yes, since the hook does the fetch after I push from my local repo
tPl0ch	but only if I run it by hand
tPl0ch	my git version is 1.7.0.5
PerlJam	wait ... that doesn't make sense.
PerlJam	The hook does not execute when you push, yes?
tPl0ch	indeed
PerlJam	and how do you know you're pushing to the right remote?
tPl0ch	git remote show origin
Ilari	tPl0ch: And how do you tell it doesn't execute?
tPl0ch	Ilari, I added a 'touch PLEASE_RUN' to post-receive
PerlJam	tPl0ch: hrm.  I think this is one of those "devil in the details" kind of things.  If you could generate a transcript of everything you did to make sure the script was executable, that you were pushign to the right remote, and that the remote isn't executing the post-receive, they might help
bob_f	Ilari: Hi, just got back from lunch and been poking a bit more.
bob_f	Ilari: If I unset ssh-agent env vars, doing "git clone git@gitolite:gitolite-admin" works - you see it is asking me for the passphrase for ~/.ssh/bob and not ~/.ssh/id_rsa

Popular searches: minion pro texworks mac   openttd roadtrain   Reconfiguring network interfaces   

Generated by irclog2html.pl 2.1mg by Jeff Waugh - return