IRCloggy #git 2007-07-15

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.

2007-07-15

cehteh eh00:00
use the admin/git-mrproper.sh i made00:01
Plouj did I say something that doesn't make sense?00:01
oh yeah, I forgot00:01
I keep expecting a built-in command for that00:01
cehteh maybe there is00:01
usually git wont touch untracked files00:02
-f00:02
Proceed even if the index or the working tree differs from HEAD. This is used to throw away local changes.00:02
well .. this still preserves untracked files00:02
which is what i would expect :)00:03
deleting things which are not in revision control is very evil imo00:03
no chance to get them back00:03
Plouj well they appear because I was checking out a different branch00:04
from a different repo (yours)00:04
cehteh yes .. not tracked there i guess00:04
Plouj and I decided to just rm manually00:04
cehteh heh00:04
or that00:04
Plouj it's just that I had src/proc and the scons script was building all of that00:04
which scared me00:04
I realize now that it's from your repo00:04
and I haven't looked at your repo yet00:05
cehteh nope thats from ichthyo repo00:05
Plouj humm00:05
must be not the scons branch then00:05
→ spearce joined00:06
drizzd git clean?00:10
→ danology joined00:11
cehteh thats new?00:12
drizzd I don't think untracked files can appear from checking out a different branch00:12
spearce they can if .gitignore changed. ;-)00:12
drizzd I don't know if it's new. It's there and it removes untracked files afaict00:12
huh, didn't think of that00:13
cehteh well we solved the case00:13
spearce git-clean has been around for a while. so its not exactly new.00:13
drizzd but they don't magically appear, they're just recognized by git status00:13
cehteh git ls-files -o | xargs rm -f << is basically what i use to cleanup .. maybe git clean is similar00:13
robin hi spearce00:13
spearce that should be what `git clean -x` does cehteh.00:13
evening robin.00:14
cehteh ok00:14
drizzd cehteh: so other == untracked?00:14
cehteh i think yes .. so far it did no harm this way00:14
drizzd git clean doesn't clean ignored files, whereas git ls-files -o lists ignored files00:14
ok, makes sense00:15
robin spearce: could you push my master to repo.or.cz. It's David commit plus som minor fixes of mine00:15
drizzd I just wonder why they're called 'other' then. Untracked would be much clearer.00:15
spearce robin: done00:15
Plouj I see00:16
cehteh next time i try git clean -xX00:16
robin spearce: that was a quick review :)00:16
spearce robin: i figure you liked what you had, it was a fast-forward, so i shoved it out anyway.00:17
robin yes, I think it's ok.00:20
I learnt how to draw graphics in tables too :)00:22
not in master yet, though00:22
spearce oooh. swt usually isn't flexible like that. that must be some icky voodo.00:22
robin it is00:25
kind of kludge interface that came with eclipse 3.200:25
kludgy00:25
→ gordonh joined00:25
spearce heh. funny how often when i think of swt my mind then goes to "kludge". :)00:25
robin wierd that Event.index sometimes means row and sometime column00:26
spearce: hehe :)00:26
← gordonh left00:27
robin found out that a treeview has quadratic performance ... :(00:28
seems a table doesn't00:28
spearce feh. nice features.00:28
robin after 10000 items performance becomes noticable, after 40000 is goes out the window00:29
the kernel has about 50000 commits00:29
it just hangs for ~40 seconds00:30
spearce so its like the problems pane when there's >20,000 errors shown in it and you do Project->Clean? ;-)00:31
hmm. actually now i wonder if the reason why that pane sucks so badly is because its actually a tree view.00:31
robin only for a large number of items00:32
it sucks for other reasons too :)00:32
spearce well my projects at work tend to have a lot of warnings and/or errors. like 20,000+. :)00:33
robin gitk and qgit suck pretty good on large projects too00:33
spearce: and you don't filter them?00:33
but yes, 20000 items is noticeable00:33
spearce filtering the problems pane actually makes the damn thing perform so badly that its not worth it.00:33
when eclipse decides to update that pane the mouse stops responding on windows.00:34
for 10 minutes at a time. :)00:34
robin 20000 items takes 6 seconds in nowhereland00:34
→ clsdaniel joined00:34
spearce the only gitk suckage i've really seen is when you ask it to show a massive diff. otherwise it handles my large and complex graphs reasonably well.00:34
robin it takes time to load on a large project00:35
spearce include a --not. ;-)00:35
robin maybe I just have a slow disk then00:36
loading commits .. spin spin00:36
with a hot cache it's better00:37
unfortuntale I won't be able to make it as fast as those anyway00:38
cehteh considered to add memcached to gitweb?00:38
would be a low hanging fruit i think00:38
robin maybe I just have too little memory00:41
← spuk left00:43
→ dash_ joined01:00
← wincent left01:01
← z3ro_ left01:05
← QuercusMax left01:06
← lyakh left01:08
→ z3ro joined01:11
→ QuercusMax joined01:12
→ jcollie joined01:51
→ brothers joined01:58
→ kanru joined02:01
← tchan left02:09
→ aruiz joined02:10
→ tchan joined02:11
→ spearce_ joined02:23
← spearce left02:23
← QuercusMax left02:23
← tcoppi left02:33
→ tcoppi joined02:34
← madewokherd left02:36
→ madewokherd joined02:49
← z3ro left02:54
→ z3ro joined02:55
← jasam left02:56
→ weitzman joined02:56
→ aroben joined03:02
→ spuk joined03:10
→ spearce joined03:14
← spearce_ left03:14
← tcoppi left03:20
→ tcoppi joined03:20
→ loops joined03:24
→ doublec joined03:36
← weitzman left03:36
→ doublec_ joined03:40
← Myrizio left03:41
→ mgrimes joined03:51
← doublec left03:59
← doublec_ left04:00
→ rkaway joined04:27
← cortana left05:04
→ jrockway joined05:04
→ tcoppi_laptop joined05:14
← mgrimes left05:17
← duncanm left05:19
← rlb3_ left05:20
→ diacritical joined05:33
diacritical spearce http://www.selenic.com/mercurial/bts/issue32605:34
when editing my own files, and when working with revisions which are timestamped, the timestamp is extremely useful05:35
"i know i made a change betweeen the time my dog died and when i scored with that redhead"05:35
the problem is when i import an old package to git, having 100s of files, some timestamped in the 1990s, onto a branch. then i do a checkout and get a directory with all files haveing the same time05:37
all that information of when the "contents last changed" is gone.05:37
though no content has changed05:37
i just switched between branches05:38
the `make' argument wont hold here. no file contents chaged, just the timestamps got updated05:38
so all targets will be unnecessarily remade05:38
(assuming i had to do a `make' i.e.)05:39
spearce, but at least that url documents what i wanted to say the last time i came here and was chased off by randal05:40
spearce, i also have an argument for the idea you dismissed about a mechanism that worked for "initial import"05:45
my god theres a huge thread on this on the mailing list05:49
loops diacritical, when switching branches in git, only files where _content_ is changed will have their timestamps changed.05:59
what this means.. is that you can quickly do an 'ls -ltr' (or whatever) to see the files that changed between branches06:00
you lose this ability if you always restore old timestamps.06:00
As gitster pointed out (on the mailing list i think) there are places where showing the original timestamp make sense (eg. when creating a tar archive perhaps)06:01
but always restoring old timestamps is actually going to get in the way of many workflows06:02
diacritical loops, i want a "clean view" i cant get it06:02
i remove all files and want the repository view. that should include content-modfied times06:02
loops diacritical, i hear you. all i'm suggesting is that you hear what the "other side" wants too.. and find a way to get what you want, without hurting the current work flows.06:03
diacritical i know i am talking past you but the whole opaqueness to this concern is overwhelming06:03
now i do the checkout, some files exist only on one branch,06:03
when switching between branches i lose information06:04
i have to do a targz of the working directory everytime to support my workflow if i have to use git06:04
and then restore that06:04
loops well yes, you do currently lose the original timestamp.. _however_ you gain other information as i just tried to explain06:05
those files that don't exist on the other branch.. will now have a new timestamp and are therefore very easy to spot with 'ls -ltr' for example.06:05
and from what i can gather.. more people rely on this behavior, than are hurt by it.06:05
diacritical so much for social engineering!06:05
← spearce left06:05
loops if git were to always follow your suggestion.. it would actually hurt many users workflow06:05
diacritical no no06:06
you dont get my point06:06
rcs does this with a -T option06:06
i.e. "set modified time"06:06
you dont have to force anyone to use it06:06
see i can only be forced to use git or not use git06:07
if i have to use git, i dont have the option of preserving "time of this revision" information on the files in my working directory06:07
even if i wanted it i dont have it06:08
loops well its not currently built into git.. but git is easy to extend to include this info06:08
diacritical where would it be put?06:08
loops it's just that not many people have been asking for it06:08
diacritical git is too complex to grok and i havent looked hard enough06:08
that shows that the newbie adoption has been overwhelmingly successful and in being adopted only by newbies!06:09
loops well.. one option, would simply be to hide a .metadata file inside each commit that holds this info06:09
your -T option could look it up and touch the needed files.06:09
but that's just one implementation idea.. and i'm sure the real Git gurus would have other ideas as well.06:09
diacritical so far in experimenting if git will be suitable for my personal use, i'ev had to run :r! find . -path './.git' -prune -o -path './hg' -prune -o -type f -exec ls -lt --full-time \{} \;06:10
when editing the commit file06:10
loops i still find it astounding that someone designing a content tracking system would leave out tracking revision dates!06:11
"because newbies wont need it"06:11
loops diacritical, i don't think that's the reason it was left out at all06:12
diacritical or will get confused or something06:12
loops the reason, as i understand it, was that linus firmly believes that such attribute data is a local matter, not a useful piece of distributed data.06:12
diacritical ah06:12
loops and frankly. i've been using git from the first week it was released.. and its never been a problem for my usage.06:13
diacritical the attribute data is a means to track "content last modified time" it should not useful in any sense critical to the git system in that git shouldnt rely on it for any of its internal operations06:14
loops that's not to say that it isn't a problem for you.. but Git was designed rather thoughtfully i think for what it was intended to handle.06:14
diacritical but i can understand the curency this meme holds it if linus is behind it06:14
loops ok06:16
i'm stil astounded :)06:16
loops diacritical, don't give up.. if it's something that is really important to you.. Junio and others will listen06:16
diacritical i think i've seen junio and tglx oppose this a few times since 05 itself on the list06:17
diacritical on quick searching06:17
loops but you have to understand their perspective too, for many (most?) people using git, its just never an issue.06:17
diacritical yes that is the social reasoning06:17
loops diacritical, well it's practical :)06:17
diacritical but i wouldnt want to forced to put git between me and my data ever if it didnt do that06:18
loops no sense putting effort where there isn't a demand.06:18
diacritical well it only means more effort when that demand comes to light06:18
or a clear reason for alternative :)06:19
loops btw, Junio recently said on the list that he thought your request was reasonable. at least in the one example of expecting git-archive to produce a tar volume that had original timestamps.06:19
but i believe he also intimated he wouldn't be the one doing the coding to implement it :)06:19
actually.. i think he may have said that here in IRC rather than on the list06:20
diacritical i dont see junio on the yakov thread. that list is too chatty, it scares me06:20
i subscribed last week on gmane, after 100 msgs in one day i unsuscribed06:20
then today after a week it is 700 messages06:20
so i guess that makes ~100 msgs a day06:21
loops yeah.. the list has gotten pretty chatty.. still nothing like the kernel list or a few others06:21
diacritical sorry im just dumb to handle that much info! :)06:21
loops i unsubscribed for a while myself and have been surprised by the volume since returning06:21
→ robewald|work joined06:22
loops have to run.. cheers diacritical06:24
diacritical later06:24
← diacritical left06:28
← DrNick left06:29
← aroben left06:42
→ mithro joined07:05
← jbowes left07:10
← spuk left07:14
→ mgrimes joined07:34
Roomster → Romster07:36
← sewall left08:07
← dduncan left08:08
← madewokherd left08:11
→ evilchelu joined08:28
→ Pistahh joined08:31
→ aroben joined08:40
← janm left08:41
→ ponto joined08:42
ponto Hi, I am trying to get the pahaole project with the following command: git clone http://git.kernel.org/?p=linux/kernel/git/acme/pahole.git Is this the right one?08:43
→ ofri joined08:43
ponto With this one i get an error: git-clone: line 381: cd: /home/ponto/tmp/pahole/.git/refs/remotes/origin: No such file or directory08:44
aeruder ponto: no, that is not right08:48
they give you the urls right at the top of that page (use your web browser)08:48
i'd recommend08:48
git clone git://git.kernel.org/pub/scm/linux/kernel/git/acme/pahole.git08:48
ponto aeruder: thanks. that works08:49
→ doublec joined08:52
→ cortana joined08:52
← aroben left08:56
→ lyakh joined08:57
← evilchelu left08:59
→ Malesca joined09:00
← aruiz left09:02
→ Yuuhi joined09:03
→ nashdj joined09:14
→ nud joined09:25
← cortana left09:29
→ gordonh joined09:39
Malesca If you want to change the last commit message (as opposed to change what's files are committed), is "git reset HEAD^" and a re-commit your best bet?09:45
← gordonh left09:47
albertito Malesca: try git commit --amend, it can "amend" the previous patch, and you can re-edit the commit message09:49
→ gordonh joined09:49
Malesca Ah, great. I've used amend, but it seemed I could not change the message. I suppose I just can't use the "m" flag but will instead have the message opened in my $EDITOR?09:50
← nashdj left09:55
← mgrimes left09:56
albertito Malesca: I haven't actually tried because I often use the "git citool" interface, or just use $EDITOR, but you can always try =)09:57
→ standel joined10:04
← standel left10:06
← gordonh left10:12
→ doublec_ joined10:13
← doublec left10:13
→ standel joined10:18
← standel left10:18
→ goloo joined10:29
→ janm joined10:34
← janm left10:35
→ janm joined10:39
→ chris2 joined10:40
→ doublec joined10:44
← doublec_ left10:45
← janm left10:46
→ QuercusMax joined10:55
→ nashdj joined10:58
→ asksh joined11:04
lizac how can I do a shallow clone of just one head?11:07
mugwump oh, fetch --depth 1 works11:09
lizac ah, I see. got confused by the man page, thx11:10
→ ask__ joined11:14
→ davi joined11:17
robin ah, my first real bisect :)11:21
← doublec left11:22
← asksh left11:23
madduck why does `git-check-ref-format master` return 1 while `git-check-ref-format upstream/master` return 0?11:25
← QuercusMax left11:25
→ QuercusMax joined11:28
loops madduck, the man page says that only heads and tags are checked. (which would exclude remote branches)11:28
madduck, hmm.. since it's only checking format i guess that doesn't really matter :o/11:32
madduck, "master" is not considered a valid ref format because it only has a single component.. the code demands at least heads/blah, not just blah11:37
madduck ic11:38
loops would be nice if the man page mentioned that.11:39
→ ShadeHawk joined11:42
← kanru left11:43
→ nikodemus joined11:47
→ janm joined12:21
← drizzd left12:21
← ShadeHawk left12:22
→ evilchelu joined12:31
← janm left12:32
→ weitzman joined12:49
→ janm joined12:50
← goloo left12:54
→ weitzman_ joined12:55
→ gordonh joined12:56
← weitzman left13:11
weitzman_ → weitzman13:11
→ standel joined13:19
chris2 this new $Id$ stuff for git, can it also do the changes for each file? (iirc $RevLog$ in cvs)13:23
← Pistahh left13:31
← chris2 left13:38
→ Lash| joined13:40
kblin git has $Id$ support now?13:44
LotR yuck13:44
loops yup.. and yuck indeed.13:46
← weitzman left13:46
← nud left13:46
→ weitzman joined13:49
→ drizzd joined13:50
← gordonh left13:59
→ nud joined14:01
→ lichtblau joined14:03
← weitzman left14:16
→ Pistahh joined14:19
← mithro left14:22
→ jasam joined14:27
→ Gitzilla joined14:38
→ madewokherd joined14:49
→ GyrosGeier joined14:58
→ orospakr joined15:00
← orospakr left15:07
→ lcapitulino joined15:33
← nud left15:41
→ nud_ joined15:41
← standel left15:49
→ Randal joined15:55
← danology left16:13
→ danology joined16:14
← ponto left16:18
→ fhobia joined16:23
← nashdj left16:25
→ standel joined16:28
← standel left16:28
← nud_ left16:39
→ nud joined16:48
→ DrNick joined16:50
→ robfitz joined16:55
→ duncanm joined17:22
duncanm is there a way to 'merge' git repositories?17:22
i guess another thing i could do is wait for 1.5.3 and use subprojects17:22
loops duncanm, simply fetch the branches you want from the other repo, and then you can merge as desired.17:23
this was actually done a few times in the history of the Git project itself.. for instance gitk started out as a completely separate project and was merged into Git keeping all of its history17:24
duncanm right17:24
so what is it that i do?17:25
i make a new repo, and git fetch git://existing-repo?17:25
and then git add?17:25
loops well.. start out with a clone of the first repo..17:25
→ robfitz_ joined17:26
loops git fetch a branch from the second repo into it..17:26
duncanm what do you mean 'into' it?17:26
loops 1 sec17:26
from the first repo, do: git fetch git://repo2 repo2branchname:newbranchname17:28
jasam when you said that you needed a way to 'merge' git repositories, we understood that you had two different repositories, and "merge" is the way to merge one of them into the other17:28
loops this makes a copy of the repo2 branch inside of your repo1.17:29
you can then just treat this as any other branch... it doesn't matter that it started out in a different repo/project17:29
Randal I'm unclear on what you want to accomplish by "merging" them.17:30
duncanm i've been using git a fair bit, but so far,i think i've just been using it like a glorified SVN17:30
loops Randal, he may want to achieve the same thing Linus did by merging the gitk project into git proper.17:30
duncanm cuz i don't know how to branch and merge17:30
Randal yeah, so there's a commit somewhere that points down one leg for all of gitk history and the other leg for all of git history17:31
loops yes17:31
Randal prior to that commit, you'll see only one or the other in your repo17:31
is that what you want, duncanm?17:31
duncanm Randal: something like that17:31
Randal if not, then can you elaborate?17:31
duncanm i have 2 repos, with 2 histories17:31
Randal this is more for my education than being able to help, by the way, so ignore me if you just want to jump to the result.17:32
:)17:32
duncanm i want to end up with one repo, which contains all history (of my 2 repos)17:32
Randal including today, when you "merged" them?17:32
duncanm i have foo.git with files 'a', 'b', and 'c'17:32
Randal and two separate histories prior to that.17:33
duncanm and bar.git with files 'x', 'y' and 'z'17:33
i'd like to get a baz.git with files 'a' thru 'c' and 'x' thru 'z'17:33
Randal and if anyone pulls from you after today, they get all six files.17:33
and all the history of both projects?17:33
loops duncanm, then yes.. you want to merge then..17:33
duncanm right17:33
Randal Yeah, then you want the git/gitk style of blending.17:33
loops first step, get the branches that currently exist in separate repos, into a single repo, then use git-merge to combine them.17:34
duncanm and i'd also get a extra commit that says 'i merged on this day, etc etc'17:34
i don't really need that commit, but it doesn't hurt17:34
Randal not "extra"17:34
*the* commit that represents the merge17:34
duncanm well, if i wait for 1.5.3, then i can use subprojects, right?17:34
Randal if you think you ever want your other project standalone again, you should wait yes.17:34
duncanm but the files won't end up in one dir17:34
aeruder subprojects don't combine into a single sdir17:34
duncanm right17:35
loops duncanm, sure.. although that's meant more for projects which are maintained independently17:35
duncanm i understand that17:35
i dunno17:35
i'm writing scheme, and while there's a module system, there isn't really a LIBRARY_PATH type of thing17:35
so there isn't a simple way to loading up a 'library'17:36
Randal I'm not up on what submodules added yet17:36
I need to read. :)17:36
duncanm but if i merge all my files into one dir, then it'll be easier to do that17:36
aeruder Randal: its similar to svn:externals if you are familiar with that17:36
Randal a bit17:36
duncanm what's the timeline on 1.5.3?17:36
Randal Just as I started to learn about svn, git came along, so I got to avoid all of that except "how to check out" :)17:37
← robfitz left17:38
Randal I'm tracking the main branch, and I have git-submodule now17:39
1.5.3.rc117:39
duncanm i have a checkout of the main branch tree, but i don't install from it17:39
i was reading junio's blog, i didn't know he's japanese?17:39
his name doesn't sound japanese17:40
Randal "master" is hosed so rarely that I don't mind running it live17:40
aeruder i'm not sure the user side of submodules is really done yet...17:40
Randal "next" is a bit more freaky17:40
Pistahh duncanm: where is his blog?17:40
duncanm http://gitster.livejournal.com/17:40
→ spuk joined17:41
duncanm Hamano is a japanese last name....17:41
hmm17:41
mugwump yeah, I'd guessed hispanic from the first name until now :)17:42
duncanm Randal: git-stash in 1.5.3 seems pretty sweet17:42
aeruder git rebase -i is sweet too17:43
duncanm oh, i don't know what that does17:43
maybe this is a good time for me to learn how all this branching and merging works17:44
i always thought, when i use a branch to develop a feature, when i merge it back to main/master, i just wanna see one commit, instead of all the trials and errors i made on the branch17:45
for a while, i use guilt instead of branching17:45
loops duncanm, its not that hard.. there's some good help in the tutorials.. setting up a repo just to play in is a good way to try things out risk free17:45
duncanm i think maybe it's not so much the command set that's confusing to me; but the workflow17:46
→ kanru joined17:46
loops well there are quite a few different ways to use branches.. but think of Git. how there is a master, next, pu, html and man branches17:47
master being mainline.. and next getting the commits that will go into the new version17:48
just a way to keep those changes separate, even though they're all part of the same project17:48
aeruder duncanm: generaly unless its a small change, you want all the errors17:49
or at least all the history17:49
you can however merge just the diff, doing something like git merge --squash17:49
which just commits the diff17:49
duncanm ahh17:49
yeah, that's what i've been looking for17:49
it wasn't in the command set in the earlier, more 'hard core' days of git17:49
aeruder yea, it is somewhat recent17:49
duncanm aeruder: i prefer that the history shows only the 'good' way of how something was done17:50
it makes really the logs and history easier17:51
aeruder it does sometimes, like i said, depends how serious of a branch it is17:51
duncanm right17:51
loops in earlier days you'd just have to do a diff of multiple commits and apply it as a unit to new branch17:51
aeruder loops: you can still do that too17:51
loops sure..17:51
aeruder there are lots of different ways to go about it17:51
duncanm oh, someone was telling me that in darcs, there's a way to select specific sections of a diff to commit17:51
aeruder even rebase -i17:51
duncanm: git add -i or git-gui will help you with that17:52
duncanm i thought that's kinda nice, so a lot of times, i commit typos/stylistic changes alongside with my more significant patches17:52
oh, interesting17:53
it'd be cool if there's a way to do that in emacs17:53
Stage this hunk [y/n/a/d/j/J/?]?17:54
hmm, that's kinda too much...17:54
→ aroben joined17:54
duncanm hmm, the man page says there's also a k/K answer17:55
← aroben left17:56
→ aroben joined17:56
→ devogon joined18:01
→ cm_ joined18:01
aeruder duncanm: use git-gui then18:11
→ goloo joined18:19
← kanru left18:21
← cmarcelo left18:29
duncanm i guess my tk apps could look prettier18:39
→ git-user joined18:46
← drizzd left18:47
duncanm w/guess/wish/18:47
aeruder duncanm: cvs of tk apparently looks much better18:47
duncanm heh18:48
git-user is there a way to stop git listing all untracked files when one commits? i know i could use .gitignore, but adding all files therein, even with wildcards would be a nightmare. would be nice if git coudl just be told not to list them in the comments of the commit template18:48
duncanm the git-status manpage is a bit bare18:49
aeruder git-user: i doubt there's a way, you'd be much better putting them in your .gitignore for everyone's sake18:50
loops git-user, why so many untracked files?18:50
duncanm loops: do you know when 1.5.3 will be out?18:50
how long is the rc cycle usually?18:51
aeruder i'd imagine the answer to that is 'when it is out'18:51
:)18:51
duncanm hmm18:51
loops duncanm, sorry i don't. although i read on Junio's blog of 2 days ago "I fear that 1.5.3 is not imminent"18:51
duncanm oh18:51
did he say that?18:51
loops http://gitster.livejournal.com/18:51
duncanm oh yeah18:51
"Pushed out 1.5.2.4, as I fear that 1.5.3 is not imminent and there are enough fixes worth getting in the hands of people who want extra stability. Rather, illusion of extra stability ;-)"18:52
← git-user left18:52
duncanm sigh18:52
actually, interaction between git-stash and guilt would be nice18:52
not-here → still-not-here18:54
→ alley_cat joined18:58
← brothers left19:03
← Lash| left19:06
→ Eludias joined19:15
→ standel joined19:28
← standel left19:29
← aroben left19:52
→ tochiro joined19:59
→ cort joined20:01
← Malesca left20:02
→ duncanmv joined20:04
→ xl0 joined20:12
xl0 Is there already some git porcelain tool available to checkout a commit, edit it, and merge all the later commits on top of it? I know it could be done with manually creating out a branch from the commit and rebasing, but maybe there is already some tool available for the task?20:15
← ofri left20:18
→ Yuuhi` joined20:19
loops xl0, that sounds very close to what stacked git (stgit) and guilt are designed to do. Don't think there is anything in mainline git to do that in less steps than you describe.20:20
xl0 Ok. branch and rebase is not that hard after all. Thank you.20:21
fhobia i made a .git repos in /a ...and i have tracked files under /a/b ...is there a way i can move the .git to b ?20:23
loops fhobia, you no longer want git to manage the /a directory.. only /a/b ?20:25
fhobia yeah20:26
cehteh could imagine some dirty tricks to do that20:26
fhobia looking at rebase..20:27
oh, i could also move everything right ?20:27
then do a monster commit20:27
8)20:27
loops yes.. you need to move everything out of /a/b, into /a (using "git mv")20:27
and then once all the files are in /a.. you move the entire directory.. (all files and the .git subdirectory) into /a/b20:28
that would tell git that there is no longer a sub directory b/20:29
fhobia also a has nothing except b20:29
loops sure.. makes things easier20:29
duncanmv has a question about git-svn.20:29
← xl0 left20:29
loops duncanmv, might as well ask.. maybe somebody who knows something is listening :)20:29
duncanmv When I do git-svn fetch, it updates the git-svn remote, then I merge git-svn on master20:30
fhobia oh, rebase has nothing to do with this i don't think20:30
loops fhobia, nope.. you're not changing the history..20:30
duncanmv I have another local branch, which I merge from master (or directly from git-svn)20:30
loops fhobia, you'll need to create one (or more) commits that move the files out of b/20:30
duncanmv but when I do git-svn dcommit, which branch does it use as source ?20:30
← cm_ left20:30
← fhobia left20:32
loops fhobia, well actually.. now that you say that.. you maybe _could_ use git rebase to go back and fix up all the original commits (moving the files to the proper place)20:33
or perhaps even trying git-filter-branch (but i've never tried it myself)20:33
→ fhobia joined20:34
→ MrPloddy29 joined20:34
loops duncanm, it uses the currently checked out branch20:35
oops.. duncanmv20:35
duncanmv so I dont need to merge from master to git-svn before dcommit ?20:35
duncanm haha20:35
more duncans!20:35
duncanm: don't you use git-svn rebase?20:35
duncanmv duncanm: hi, I know you, we are collegues, I met you in boston didn't I?20:35
duncanm eek, even i got it wrong ;-)20:35
oooh!20:35
yeah, i do know you20:35
cool20:36
duncanmv I think git-svn rebase is obsolete, git 1.5 does not have it, but only fetch20:36
duncanm that's not true20:36
← Yuuhi left20:36
loops 1.5.3 still has rebase20:36
duncanm i think git-svn rebase is the recommanded usage20:37
okay, gotta go20:37
loops bye duncan (m)20:37
duncanmv duncanm: mine does not recognize rebase20:37
loops hmm.. is it still in the man page?20:38
← MrPloddy29 left20:38
→ MrPloddy29 joined20:39
→ brothers joined20:44
fhobia hmm git mv is buggy huh? sometimes it deletes and copies files rather than moving...why is that ?20:44
loops fhobia, think you ducked out just as i was saying that you might be able to use git-rebase or git-rewrite-branch to fixup old commits, moving the files to the place you really want them20:47
it would depend how complex your history is to decide if it was worth it20:47
as for "git mv" what version of git are you using? what problem are you seeing?20:48
→ orospakr joined20:48
← GyrosGeier left20:50
fhobia ah, thanks loops20:50
i'm using 1.5.1.320:50
instead of saying moved, it says deleted for some files and then later copied20:51
insead of saying renamed or whatever20:51
i don't know how to reproduce it since i was doing files in bulk20:51
loops oh.. you mean just in the commit message comments?20:51
fhobia yeah20:51
it also said like: copied: file1 -> file2 copied: file1 -> file320:51
which doesn't make any sense...20:52
loops well, internally git doesn't have any structure to represent moves.20:52
fhobia hmm i see, i did a -n to see what it would do...it just does add and delete under the hood it seems20:53
loops yes.. under the hood.. there is nothing that remember this file -went-> there20:53
git uses some heuristics to offer its best guess at what renames, and copies etc happened in a given commit.20:54
perhaps it's bad advice to suggest to you to move them all in one commit..20:55
fhobia looks like it may work..20:55
loops well it will work.. but it might make your history a bit hard to read.20:56
← lcapitulino left20:57
← jrockway left20:57
→ jrockway joined20:58
Tali fhobia: this happens if you have multiple identical files and move them. Then git cannot deduce which got moved where.21:00
everything still works, it just does not look that nice...21:01
→ gordonh joined21:06
fhobia ah ok21:06
Tali: that makes sense...those files that didn't work were all empty21:06
← Eludias left21:08
→ Eludias joined21:08
Randal there's only one "empty" file in the repo. :)21:12
empty blob, that is21:12
if you have many copies of that, and they get renamed, it's really hard to even tell what happened. :)21:13
Hmm. That could be a good trivia question. "What is the sha1 of the empty blob?"21:13
fhobia :-) well, that isn't so bad for my case21:13
i did git-push to remote server and now the remote server shows the patch in the log but hasn't committed it?21:14
← gordonh left21:14
Randal if it's in the log, it's committed.21:14
are you saying not reflected in the working tree?21:14
fhobia yeah, i moved a ton of files in my local repos ...the did git push to remote repos..and no files were moved21:15
Randal right21:15
don't push to a non-bare repo21:15
that's the first thing to learn. :)21:15
fhobia lol21:15
Randal but you can go into that repo and say "git-reset --hard HEAD"21:15
I hope you didn't have any local edits21:15
fhobia nope21:15
Randal if it's just a checkout that's tracking a head, then it's not bad21:15
but you'll need to add a trigger to do that git-reset for you when you push21:16
or take my first advice: don't push to a non-bare repo!21:16
fhobia ok, that reset worked...what did that do? it didn't seem to do anything across the wire21:16
Randal right. it's making the local tree reflect the local commit21:16
fhobia yeah, i'll take the first advice rom now on21:16
Randal pushing doesn't update the tree. just the objects21:17
fhobia doh!21:17
Randal objects and refs21:17
→ gordonh joined21:17
Randal so then you had to say "now make the tree match this"21:17
this is By Design21:17
← gordonh left21:18
fhobia thanks loops, Randal, Tali :)21:22
darn, my server has git 1.121:23
and i use git 1.5 locally21:23
aeruder holy moley, 1.121:23
that's not even usable, heck i barely consider 1.4 usable21:23
→ gordonh joined21:23
fhobia i hope the underlying protocol is the same..21:24
Randal your server should be updated immediately21:26
before you get something broken21:26
← meyering left21:29
→ rtmfd_icbm joined21:32
→ sewall joined21:32
← alley_cat left21:40
← gordonh left21:47
→ kristian_ joined22:02
kristian_ I've seen some posts about a eclipse plugin for git but I've not been able to get a clue as to how complete it is or not. Anyone that knows more about such a project?22:03
Randal ... http://repo.or.cz/w/egit.git22:03
that was 4th on a yahoo search22:04
there are these... you know... search engines... :)22:04
kristian_ I've seen it but it doesn't directly say which state the plugin is in22:04
Randal: thanks anyway22:04
Randal what do you mean "state"?22:04
kristian_ Randal: does it implement the necessary features to be useful for example22:05
Randal: is it reliable or buggy as hell22:05
Randal A quick email to the authors would reveal that22:06
← Yuuhi` left22:06
Randal given that the most recent tag is v0.2.2 should also be a clue22:07
and that was 2 months ago22:07
although the most recent commit was 46 hours ago22:07
again, all gleamable from that web page22:07
gleem?22:07
no - gleam22:08
→ baggito joined22:09
kristian_ I've was hoping for that someone had tried it or was using it, I guess I'll have to test it myself to find out22:09
Randal Yeah sorry I never saw the need for Eclipse myself.22:09
kristian_ I used to use gedit and a terminal for all my coding but when I got used to eclipse at work I got kinda sold out on it22:10
tje22:10
there is just one thing that sucks and that's that it's written in java22:10
Randal Yeah, that'd almost be a deal killer for me.22:10
kristian_ it easily takes 300mb ram22:10
Randal although you can claim that my IDE of choice (GNU Emacs) is roughly the same. :)22:11
kristian_ ;)22:11
i use emacs for all config files and other edits so to speak, but I guess I'm to lazy to learn to use it as an ide22:12
Randal well - I never learned to use "a debugger". I debug perl code by adding print until it works.22:13
old school, I guess.22:13
kristian_ I see, with perl I guess that's my approach also22:13
in bigger c++ projects I do usally use gdb and a set of debug classes22:13
← kristian_ left22:14
→ doublec joined22:18
← lyakh left22:28
→ PugMajere joined22:31
← nikodemus left22:35
robin Randal: I didn't know adding print's will make the code work :)22:38
Randal it seems odd when you first hear it, but it works.22:40
been coding that way for 35 years. :)22:40
← Eludias left22:42
→ hein joined22:43
← davi left22:43
→ branstrom joined22:51
← duncanmv left22:57
← Sho_ left22:59
hein → Sho_22:59
→ Myrizio joined23:05
branstrom Hmm, can anybody help a Linux newb? I want to color the lines that I write myself, in the terminal (PuTTY, but that shouldn't matter cuz the serverside controls the colors)... Is that possible?23:06
Randal crayons? :)23:07
aeruder branstrom: it kind of depends on the shell you're using23:08
also, it depends what you really want23:09
you can change the default color of your terminal, which would include what you type23:09
branstrom I found bashish. Using bash now.23:09
http://bashish.sourceforge.net/23:09
cort cool23:09
Randal is "branstrom" a name, or a play on "Brainstorm"?23:09
branstrom "Bränström" is my last name.23:10
Randal ok23:10
branstrom But it looks kinda like "brainstorm", which is a plus :p23:10
Randal yeah - that's what I wondered23:10
← doublec left23:10
← goloo left23:11
→ doublec joined23:11
→ gordonh joined23:14
→ rlb3 joined23:21
← janm left23:40
← orospakr left23:57

Logs Search ←Prev date Next date→ Channels Documentation