IRCloggy #git 2007-05-22

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.

2007-05-22

→ prxq_ joined00:04
→ gitte joined00:12
→ lu_zero joined00:35
→ meyering joined00:35
→ kblin joined00:35
→ kblin joined00:51
→ orospakr joined01:07
→ lu_zero joined01:23
→ meyering joined01:23
→ aroben joined01:31
→ aroben joined01:34
→ DrNick joined01:42
→ spearce joined03:09
→ sgrimm_ joined03:58
→ devogon joined04:18
dwmw2_BOS → dwmw2_gone04:26
→ yorgen15 joined04:26
→ rkaway joined04:27
→ Romster joined04:30
→ DrNick2 joined04:57
aeruder__ → aeruder05:00
DrNick2 → DrNick05:02
→ sgrimm joined05:13
→ kanru joined05:45
→ Tv joined05:49
→ aroben joined06:06
→ aroben joined06:06
→ etnt joined06:22
→ noss joined06:58
→ raj joined07:28
← raj left07:28
→ amitg_ joined07:28
← dduncan left07:39
→ agoode joined08:08
→ janm joined08:11
→ fultilt joined08:15
prxq_ → prxq08:15
→ Oejet joined08:29
→ gitte joined09:08
→ segher_ joined09:13
→ GyrosGeier joined09:26
buggs gitster, since i copied the git repository to an ext4dev fs i no longer experience bit flips09:37
so that might mean it was the disk or XFS09:37
http://bugs.archlinux.org/task/651409:38
wildfire hmm - when I do a 'git pull' on a tracking branch I get a merge conflict; should I setup a tracking branch especially? Or should a branch with no changes from upstream simply fastforward?09:46
gitte wildfire: usually, you don't pull into a tracking branch. You don't even check it out, but branch from it.09:46
Thus, you should always fetch it, not pull it.09:46
"git remote add" will set that up for you (given a recent Git)09:47
wildfire okay, I'm confused.09:47
I did: git clone <url>09:48
then: git branch master origin/master09:48
I haven't done anything on my master branch09:48
but I want to do a 'git pull' which should fastforward09:48
but I get a conflict instead09:48
gitte Can you visualise it? "gitk master FETCH_HEAD" should show you where the two branches diverge.09:49
wildfire hm, that seems to display all the entries .. should that be "gitk master..FETCH_HEAD" ?09:50
kha wildfire: FETCH_HEAD..master, I think09:51
wildfire: To see what's on your master, but not in the branch you merge.09:51
wildfire ahh - it looks like upstream has re-based09:52
will a 'git pull master:master+' fix it?09:52
kha wildfire: Then maybe "git-reset --hard" if you really have no local changes.09:52
wildfire it looks like *everything* has changed, so I want to throw away what I have and just bring in the remote master; I've tried the reset hard many times and it seems to be conflicting09:53
gitte git reset --hard FETCH_HEAD09:54
kha wildfire: A recent enough git will have put plusses in the fetch lines in the config file.09:55
gitte (But if everything is different, it was no rebase. Next time you see that maintainer, don't forget your bat)09:55
kha wildfire: So just "git fetch", followed by "git reset --hard origin/master" should do it.09:55
wildfire kha, thanks, I ended up doing (bit branch newmaster origin/master) and then echoing the sha1 into the .git/refs/heads/master which worked fine10:02
kha, but I'll keep that syntax in mind, as it is shorter10:02
→ cods joined10:32
→ cods joined10:40
→ hein joined10:50
→ cods joined10:51
hein → Sho_10:52
→ cods joined11:04
→ chris2 joined11:58
→ pflanze joined12:06
pflanze Hello.12:07
Is someone working on libification of the git sources now? (there has been a Google SoC proposition)12:07
→ tahseen joined12:09
tahseen I've some tla branch where some of the timestamp are incorrect12:09
→ gitte joined12:09
tahseen when using git-archimport the order of the patches are not preserved as git-archimport sorts the patches by time before applying it (as far as I can tell)12:10
is there another way to do this ignoring the timestamp?12:10
→ mithro joined12:10
→ salty-horse joined12:34
salty-horse hi. looking at the git manpage on ubuntu, I see ".sp" at the end of paragraphs. Is this a known problem, or maybe something specific to my distro?12:36
→ lcapitulino joined12:39
dwmw2_gone → dwmw2_BOS12:59
→ GyrosGeier joined13:08
→ dwmw2_gone joined13:21
dwmw2_gone → dwmw2_BOS13:22
→ blastura joined13:29
pasky salty-horse: I don't think it's a known problem13:30
pflanze: I think the SoC project is in progress13:30
pflanze k13:30
salty-horse pasky, thanks. my question was a bit unclear. do you see the problem yourself? :)13:31
pasky I don't :)13:32
→ lcapitulino joined13:41
→ jasam joined13:48
jasam gitte, well, I just don't know how to do path_list more usable...13:57
gitte I'd probably just write two functions, append_to_path_list(string, list), and sort_path_list().13:58
s/()/(list)/13:58
jasam did you read my "reply"?13:59
gitte Hmm. I am reluctant now, since a flag to keep the path_list sorted would mean that you have to check it when accessing the list.14:01
Hmm.14:01
→ orospakr joined14:05
→ devogon joined14:06
→ tcoppi joined14:06
→ spuk- joined14:08
jasam gitte, users can take care about that14:10
gitte jasam: those are hard to spot bugs.14:11
However, I just sent you a totally untested idea of what the functions could look like.14:11
jasam what about my last suggestion?14:12
gitte I don't like that particulary, since it should be a compile time option, not a runtime option.14:12
I briefly reflected if _insert() should check the flag first, and sort if necessary, while append should clear the flag, but that is too complicated and fragile, methinks14:13
If somebody _knows_ that the list should not be sorted, there should be a chance to do so.14:13
Probably the cleanest way would be to rename path_list to sorted_path_list.14:14
Then you could easily introduce a path_list, which does what we want.14:14
jasam which of both will remove duplicates?14:17
gitte Only the sorted one.14:18
(because searching for duplicates in an unsorted list, while building it, is really really expensive)14:18
→ rphillips joined14:20
gitte Although you can do a path_list_remove_duplicates() in O(n)14:21
jasam in an unsorted list?14:22
gitte (after sorting)14:23
→ lcapitulino joined14:23
gitte Come to think of it, this should be an option to sort_path_list()14:23
jasam if you let duplicates, yes14:23
gitte append() _must_ allow for duplicates.14:23
→ orospakr_ joined14:25
← salty-horse left14:30
jasam gitte, I think is good to have both two types with no common functions, so people can see the difference easily14:32
→ krh joined14:33
jasam and choose between them14:33
→ robfitz joined14:33
tahseen is it possible to stop git-archimport from sorting the patch by time14:34
and just apply them in the order tla abrowse returns them14:35
gitte jasam: yes. you can even have "struct sorted_path_list *sort_path_list(struct path_list *);" then.14:35
jasam weird14:37
what about the original path_list?14:38
gitte I'd just rename it.14:39
It was a misnomer to begin with.14:39
jasam if both are the same14:39
I don't know14:40
gitte Well, internally they can be the same. Just do "typedef struct path_list sorted_list"14:40
jasam that didn't catch errors when mixing functions14:41
won't14:41
gitte Hmm.14:43
GyrosGeier you want C++14:43
gitte No, I don't.14:44
jasam we are just sorting strings14:44
gitte a diligent C++ programmer will write good programs. Guess three times why there are no performant C++ programs out there.14:45
GyrosGeier heh14:46
I'm just downloading Rational Purify14:46
to find the heap corruption14:46
GyrosGeier wants valgrind14:46
gitte Wonderful.14:46
I really like people who want to fix bugs with bigger bugs.14:47
→ ofri joined14:50
jasam gitte, you can use alloc_nr in path-list.c, I'm not sure, I don't really take care about how it increases14:53
I was using it as I saw it in builtin-branch.c14:53
gitte Yes, I thought about that, too.14:55
Incrementing in steps of 32 is probably stupid.14:55
→ nud joined14:59
GyrosGeier gitte, well, when you have a bug that takes a few seconds to manifest because the allocator it is busy following pointers in a linked list that someone looped back onto itself, valgrind is sort of the only option there15:01
jasam if something helps you to find bugs, why not?15:04
EvanCarroll ok after my long ordeal yesterday of complete cluelessness on patches, I tried to remedy it by reading random git docs on patches.15:08
I had some problems, I'll get back to you15:09
→ Kapitoshka joined15:13
EvanCarroll actually this time around it works great15:14
nifty15:14
← Kapitoshka left15:16
→ strangy joined15:23
jasam can I follow the history of a file using gitk?15:43
→ lcapitulino joined15:43
jasam I saw that feature in the web and it is nice to learn where things come from15:45
I was using Find, but it is a pain in a long history!15:48
gitte jasam: try this: "gitk <filename>"15:50
jasam gitte, it works, thank you, this is the first time I see the whole git-tag.sh history, it is wonderful...15:57
→ orospakr_ joined15:57
jasam I have a .txt with those SHA1 copied15:58
gitte Yeah, I like that feature myself...15:58
You can make such a file with "git rev-list git-tag.sh"...15:58
→ kanru joined16:03
→ lyakh joined16:04
jasam I need to insert HEAD before "git-tag.sh", right?16:04
gitte Oh yes. Sorry. rev-list needs explicit ranges.16:04
jasam it's ok, even better, that way I have to read the fine manual16:05
→ robinr joined16:17
→ alley_cat joined16:21
→ sgrimm joined16:24
→ ferdy joined16:24
→ aroben joined16:54
→ jeffpc joined16:59
→ dduncan joined17:46
→ l_bratch joined17:49
l_bratch With Git, how can you revert to version-xyz of something?17:49
gitte git-reset --hard17:49
l_bratch Thanks, but what if I have version 9 synced in git, and want to go all the back to version 5 for instance?17:50
gitte git reset --hard <commit-which-corresponds-to-version-5>17:51
Or do you just want to peek?17:52
→ strangy joined17:52
l_bratch Thanks gitte - that was it17:53
→ yann joined18:03
→ aroben_ joined18:06
→ tcoppi joined18:07
→ robewald|home joined18:07
→ aroben_ joined18:08
albertito Hi! I trying to import a svn repository into git using git-svn and I get the following error:18:12
Invalid filesystem path syntax: REPORT request failed on '/svn/softswitch/!svn/vcc/default': Cannot replace a directory from within at /usr/bin/git-svn line 303118:12
Any ideas on what might be going on? I'm importing branches and trunk, but not tags because it fails too (with a 'cannot lock the ref')18:13
lu_zero sorry to ask again but I couldn't find it in the wiki and seems I lost the log18:17
which is the single command to create a new remote branch from a local one?18:18
rphillips what does this mean: refs/heads/pu: not updating to non-fast forward branch pu18:21
I git pulled the master git repository for updates18:21
→ Pistahhh joined18:37
→ rkaway joined18:38
→ sgrimm_ joined18:42
→ lu_zero joined18:42
→ dst__ joined18:44
→ chris2 joined18:45
Pistahhh → Pistahh18:55
→ Karlik joined18:56
← l_bratch left18:58
Karlik hello everyone, I never use the git to get the source. I have a question: how can I get the newest tree of kernel?18:59
→ ferdy joined19:00
context karlik: kernel.org ?19:02
Karlik there I found the snapshot19:02
→ fultilt joined19:02
Karlik http://git.kernel.org/19:02
I really can't see which is the always newest repo19:03
if there is not, I will be getting downloading the snapshots19:03
gitster define "newest" -- are you interested in Linus's tree?19:04
Karlik newest: 2.6.22, after some time 2.6.23 etc.19:04
moh Karlik: you grab linus tree and then make a local branch based on a tag19:05
gitster Do you mean "newest official" or "newest work in progress that Linus integrated and in preparation for the next one"?19:05
Karlik the second definition :)19:05
gitster git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git/19:05
That's Linus's tree.19:05
Karlik thanks :)19:06
gitster "git clone $it"19:06
Karlik thank You :)19:06
albertito I "solved" the git-svn problem by running git-svn fetch -r 5650:HEAD (5650 was the commit that was failing)19:07
Karlik it solve my second problem (how to using git ;) )19:07
→ dwmw2_gone joined19:08
gitster T0:23:respawn:/sbin/getty -L ttyS0 9600 vt10019:08
OOps.19:08
http://www.kernel.org/pub/software/scm/git/docs/user-manual.html19:08
→ sgrimm joined19:08
gitster The first two chapters would help to get started with exploring Linus's tree.19:08
dwmw2_gone → dwmw2_BOS19:08
da5id albertito: You might need to run a svnadmin recover on you svn repos (be sure to back up first)19:12
albertito: you can run a svnadmin verify first19:12
albertito da5id: I don't manage the repository :S19:13
da5id albertito: crap, it sounds like it has a corrupt revision19:14
albertito da5id: if you want the information to verify that there is not a bug in git-svn, I could ask the svn repo admin to do the verify and/or give me a tarball of the repo so I can play with it myself19:15
→ sgrimm joined19:15
da5id albertito: sounds like a good plan19:16
albertito da5id: but since this seems to be working now (I guess it just skipped the weird commit and incorporated the changes into the next one), I'm not really concerned19:16
da5id: ok, I'll let you know when I have the tarball19:16
da5id: thanks a lot!19:17
→ robinr joined19:19
da5id I have just converted a local svk mirror to git - now how do I make it fetch for the original source ?19:19
→ nud_ joined19:45
→ lyakh joined19:48
→ orospakr joined19:49
albertito da5id: I'm running svnadmin verify on a local copy of the svn repository that failed; I'll let you know when it's done (if it ends today, otherwise it'll have to wait until tomorrow)20:12
→ robinr joined20:12
da5id albertito: Cool =)20:12
→ aroben joined20:15
→ ofri joined20:25
→ strangy joined20:37
albertito da5id: it was clean, I'll do a repair just in case20:42
da5id: well... it didn't take long, and didn't say anything suspicious20:42
→ meyering joined20:43
da5id albertito: wierd - I hav had similar problems with svn (not related to git), and the cause was a corrupt repository20:43
albertito da5id: well, I'll look into it sometime this week. Luckily, I could perform the import anyway, just skipping these two commits. That's good enough for me at this time =)20:46
da5id albertito: fair enough20:55
=)20:55
albertito da5id: thanks!21:00
→ amitg_ joined21:03
amitg_ I started out with 2.6.18, created a git repository out of it and made my changes on top of that..21:04
now I want to rebase my changes to the latest kernel version and I want to keep the git logs as well...whats the easy way of doing it?21:05
→ robinr joined21:06
mugwump stg init; stg uncommit -n XXX (where XXX is the number of commits going back to 2.6.18),...21:08
git remote add linus git://git.kernel.org/...21:08
git fetch linus21:08
git checkout -b mybranch linus/master21:09
stg init; stg pick patchname@oldbranch21:09
repeat for all patches on oldbranch21:10
(get the list with stg series -b oldbranch)21:10
amitg_ mugwump: thanks, will give it a shot21:11
mugwump if the conflicts become a pain be sure to set up a custom merge tool for stg21:11
I use http://utsl.gen.nz/scripts/smartmerge21:12
my that script is getting long :)21:12
→ gitte joined21:12
amitg_ nice..thanks!21:13
→ sirPalook joined21:23
sirPalook Q: after git-read-tree, the timestamps in the index file are zeroed?21:23
gitster "git-read-tree $tree" clears, "git-read-tree -m $tree" keeps, I think.21:25
amitg_ mugwump: in second step, 'remote' is not a git-command21:25
sirPalook um, read-tree couldn't get any timestamps from the database. they aren't stored21:26
gitster you have an ancient git then X-<.21:26
I think you misunderstand what the timestamps in cache entries are used for. They are to see if working tree files are changed and nothing else.21:27
mugwump amitg_: get a newer git21:27
sadly etch has been etched on ancient git21:27
amitg_ mugwump: ok21:28
sirPalook when we do update-index, the timestamp of each file is created from stat()21:28
gitster So if you read-tree, we do not get timestamp from commit and set that to cache entries. If you "read-tree -u", we know that what we wrote out to working tree matches the cache, so we record the file timestamps.21:28
sirPalook when we do read-tree, the timestamp is from where?21:28
gitster Plain straight read-tree (without -m nor -u, single tree) clears the timestamp -- you are telling the index "you do not have any particular relation with what still happen to be in working trees".21:29
Especially, if you do "rm .git/index; git-read-tree $tree".21:29
sirPalook ahh. alright. So a diff-files after a read-tree will list *all* the files21:30
gitster Yes. That's why our script Porcelains often do "update-index --refresh" after read-tree, although many codepaths now use "read-tree -m -u".21:31
... which does the equivalent internally so we do not have to do the --refresh explicitly.21:31
Are you writing your own Porcelain, or being curious about plumbing?21:31
sirPalook writting my own core ? :)21:32
→ Pistahhh joined21:44
Pistahhh → Pistahh21:53
→ yorgen15 joined22:13
robinr corecode: seems keyword expansion is off in fromcvs-konverted files22:29
off-by-one22:29
e.g. $Revision expands to 1.29 instead of 1.3022:30
it is evil and should be turned off anyway så it's not a big deal22:31
the first version is not expanded, so there is no 1.022:34
is anyone maintaining a git-repo for fromcvs. mercurial isn't working for me. I can clone, but everything else breaks22:44
as do the hg2git script22:44
context hg2git worked fine for me22:45
ive never touched the cvs portions of git :/22:45
hell ive never actually used cvs beside for checkout, im thankful for that22:46
robinr lucky bastard22:46
context ;)22:47
robinr all alternatives aren't worse though22:47
context git-svn works wonderfully22:47
robinr won't help me22:47
context heh22:47
if you need help with hg2git i can help with that22:48
robinr ConfigParser.InterpolationSyntaxError: '%' must be followed by '%' or '(', found: '%7Ecorecode/hg/fromcvs'22:49
context fromcvs looks like a question for #mercurial22:49
robinr :)22:49
context you tried git-cvsimport22:49
robinr yes, it works sometimes22:49
context o22:49
robinr fromcvs works fine22:50
context i can see that22:50
robinr it's hg giving errors, not fromcvs22:50
context is it a public repo?22:51
robinr I had one error in from, but I suspect it could have been fastimport that failed silently for some odd reason22:51
context meh nm22:51
i dont wanna touch cvs22:51
robinr no need22:51
hg clone http://ww2.fs.ei.tum.de/~corecode/hg/fromcvs/22:52
that's how I got the program22:52
→ gitte joined23:00
context robinr: clone worked fine here23:00
o23:01
program23:01
robinr: well judging by the error whatever is being interpolated has text that is bad23:02
robinr I made a new clone and managed to get my changes to it by poking around23:05
and the hg commands are working there (so far)23:05
except hg2git23:05
context heh23:05
robinr tip is 10123:09
analysing the branches...23:09
Traceback (most recent call last):23:09
File "/home/me/SW/GIT/contrib/hg-to-git/hg-to-git.py", line 110, in <module>23:09
hgchildren[parent] += ( str(cset), )23:09
KeyError: '-1'23:09
otoh, now that hg is working I may not need to convert it :)23:10
→ aroben joined23:36
→ gitter joined23:43
→ mithro joined23:47
→ Romster joined23:48

Logs Search ←Prev date Next date→ Channels Documentation