IRCloggy #git 2006-06-22

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.

2006-06-22

→ GeertB joined01:11
→ pschulz01 joined02:07
pschulz01 Is there a way to generate a report from a git archive (between commits) similar to what 'git pull' displays?02:19
eg. lines of code added/deleted per file.02:20
Or do I need to pipe output 'git diff' though some patch summary tool?02:21
DrNick git log A..B --stat02:21
although that's a diffstat per commit instead of total02:22
ah. git diff-tree A..B --stat02:24
pschulz01 Thanks.. I though I was going crazy..02:25
DrNick when in doubt, read the source02:25
pschulz01 Hmm.. I might have an old version.02:26
(git-core in ubuntu 6.06)02:27
1.1.302:27
DrNick very old02:29
→ GyrosGeier joined02:35
→ krh joined03:03
→ alp joined03:15
alp i've added a new page to the wiki that i'm fleshing out, http://git.or.cz/gitwiki/GitCommonMistakes03:24
the idea is to document botched git use and migration in the wild, feedback welcome03:24
dkowis what are the migration patterns of git in the wild?03:32
dkowis makes bad jokes03:32
dkowis s/jokes/puns03:32
dkowis → wtf03:42
wtf → ohehmeffgee03:43
ohehmeffgee → stfu03:43
stfu → dkowis03:44
→ eleusis joined04:49
→ pjz joined05:11
pjz anyone awake?05:12
situation: I did 'git clone' and then made one change that I want to keep, and then accidentally deleted a file. I haven't done a commit. How do I get the file I accidentally deleted back?05:13
→ gitster joined05:14
DrNick git checkout path05:14
pjz ah, sweet, thanks05:15
I swear I looked all over the tutorials and the manpages and etc05:16
→ krh joined05:38
Shipon → Aexoden05:40
ShadeHawk How to split commit into two?06:13
→ Eludias joined06:35
→ lukass joined07:21
→ dst_ joined07:38
→ Tv joined07:57
alley_cat I thought git checkout path/to/file would restore that?08:22
it doesn't do it here, and I can't seem to get the file out of the index either08:22
I did a quite messy cherry-pick, and there were conflicts in the file08:23
dwmw2_gone → dwmw208:32
ShadeHawk if you have dirty version of the file, try 'git checkout -f -- file'08:41
sorry, -f doesn't work with path.08:42
git update-index --remove? check update-index options.08:42
alley_cat I think I tried almost everything08:43
gave up on that cherry pick for now08:43
→ octo_ joined09:19
octo_ → octo09:21
→ alley_cat joined09:28
→ alley_cat joined09:33
→ tonyj joined09:33
→ alley_cat joined09:34
ShadeHawk Can lightweight tags point to something other than commit? git 1.3.0 doesn't warn/error out that it cannot create lightweight not-commit tag...09:44
Additionally, from documentation is not obvious that if you use -m option to provide tag comment, -a (or -s) is implied, and tag object is created.09:46
→ chris2 joined10:05
Oejet ShadeHawk: A tag can point at any type of object.10:23
→ matled joined10:31
ShadeHawk Sorry... that is just gitweb bug then, it does show only example/commit-light10:34
→ benlau joined10:37
→ ferdy joined12:05
→ pasky joined12:07
→ mchehab joined12:10
→ boto joined12:11
→ coywolf joined12:45
→ alley_cat joined13:16
→ ferdy joined13:28
← EdsipeR left13:34
→ agile joined13:37
→ mbrio joined13:38
mbrio Good morning all, I was wondering if there was a way to break up a Git repository into smaller repositories13:39
and keep their histories13:39
jdl mv13:39
Oh.13:39
I don't think so directly.13:40
The obvious answer is to clone it n-1 times and remove the unwanted bits from the clones...13:40
mbrio hrmm13:40
that could work13:41
I'll play around with it13:41
jdl The tree object encompasses the whole repository ultimately.13:41
pasky ShadeHawk: I think using stgit would really help you a lot :)13:41
jdl If it is strictly on a per-directory basis, you could conceivably locate the right tree object in the repo and...13:42
pasky mbrio: you might find cg-admin-rewritehist useful for it (it won't do it for you but will help you automate it)13:42
mbrio hahah and end up messing up my repo hahah13:42
pasky ?13:43
mbrio though it would be a good way for me to understand the internals more13:43
I'm fairly new to git13:43
but I do considerably like it13:43
at least I can screw with a local repo13:43
jdl I'm new to git too.13:43
mbrio and if I end up destroying it I atleast always have the global one13:43
jdl nods.13:44
pasky mbrio: just clone it and experiment on the cloned copy13:44
mbrio yeah13:44
pasky mbrio: note that cg-admin-rewritehist will let you create branches containing the "separated" repositories, but then it's easy to separate the disjoint branches to separate repositories13:45
mbrio I'm basically dealing with the problem I had yesterday of wanting to link external projects, now I'm breaking up each of the projects into much smaller projects13:45
pasky: that seems as though it will do what I need it to do, I'll do some research on it, I'm not using cogito but I do have it installed13:45
pasky cg-admin-rewritehist is fairly isolated so you don't need to know the rest of cogito to use it13:46
mbrio I figured before I started using an interface to git I should learn git first13:47
pasky mbrio: well cogito is rather designed for the opposite way, but whatever works for you :)13:48
jdl Heh.13:52
And how would _you_ know what Cogito is designed for, pasky? :-)13:52
pasky just guessing ;))13:53
jdl Thought that might be the case.13:53
pasky anyway the primary design goal was to just have a version control system that's easy to use13:54
mbrio Yeah I understand that it was meant so you didn't really need to know the internals of Git, but if it is being used as an interface to Git that's easier to use wouldn't you want to know what Cogito is doing under the hood so you know what to do if anything goes wrong with your repo?14:11
but then again you wouldn't need to use Cogito if you knew the internals14:12
jdl Some folks don't want to know what's going on in the Index, for example.14:12
mbrio yea14:12
and that's where the market of Cogito has been created14:12
and that's great14:12
most people don't really need to know it unless they are repo administrators14:13
→ GyrosGeier joined14:13
→ GeertB joined14:24
jdl Where are the --pretty format options documented?14:29
ShadeHawk Somewhere deep, and it seems like not all are documented. For example git-log accepts --pretty=oneline, git-diff-tree(1) talko only about raw|medium|short14:32
jdl goes grepping...14:34
jdl git-rev-list is supposed to, it sems.14:34
seems, even.14:34
git-log cheeses out and refs options from git-diff-tree and git-rev-list.14:35
Seems like we should factor those options out into their own #include-able files more.14:35
diff-tree should already be that way, IIRC.14:35
Alright. Maybe I will get to it tonight...14:36
dwmw2 → dwmw2_gone14:57
→ Tv joined15:00
→ pasky joined15:26
ShadeHawk how well git deals with hardlinks? does it create during clone or checkout hardlinks for the files which have identical contents? i.e. does it preserve hardlinks across cloning/syncing repositories and working directories?16:14
pasky you mean in the working copy or in the object store?16:19
ShadeHawk both16:26
in the repository I mean in the tree object.16:27
if it is needed.16:27
pasky huh16:28
ShadeHawk but mainly in the working copy16:29
symlinks I think are supported well.16:29
pasky well you really do not want to hardlink files with identical content in the working copy16:29
ShadeHawk unless you have COW-links (Copy-On-Write hardlinks)16:29
well, I want to preserve hardlinks ;-)16:30
pasky git can't do that as things are now16:31
ShadeHawk oh, well, it is not that important...16:32
→ chris2 joined16:40
chris2 are there any bindings to libgit.a for scripting languages?16:42
merlyn beginnings of Git.pm now16:43
pasky What's the minimal perl version we should work with?16:44
chris2: I'm working at it at this very moment :)16:44
chris2 is there anything you can show? :)16:45
chris2 'd like to do a ruby interface16:45
pasky chris2: http://pasky.or.cz/~xpasky/git-perl/Git.html (s/html/pm/ for the source)16:46
chris2: we'll primarily just call external commands now and slowly add bindings to libgit as we go16:46
but I'm already doing Git.xs for hash_object(), just for the heck of it :)16:46
chris2 i see16:47
→ pdmef joined16:51
ShadeHawk same place, same bat channel^W^W^W i.e. http://pasky.or.cz/~xpasky/git-perl/Git.xs (BTW, why didn't you do this on topic branch in your git repo, and publish via gitweb... or is it in too early stages of development)?17:06
→ GeertB joined17:07
ShadeHawk what does hash_object() do?17:07
pasky keep reloading ;)17:11
ShadeHawk is it to store object, get object from repository (cat-file), open file for reading?17:16
pasky: BTW your web server serves Git.pm as text/x-perl, so I cannot view it in my browser. Could you symlink Git.pm.txt to it, or change httpd config, please?17:18
← mbrio left17:19
pasky your browser is broken17:22
but I created the symlink17:23
ShadeHawk thanks.17:24
chris2 tries to make a dynamically linked libgit, but always gets errors due to undefined main17:24
ShadeHawk there was some patch on the list to make libgit shared library sometime ago17:25
alp i have patches for libgit if your're interested17:25
chris2 i just wonder where main is referenced17:25
dst_ It was my patch, an old version at available at http://stosberg.net/pub/17:26
ShadeHawk pasky: why do you use CamelCase keys for option names?17:26
dst_ I also have the first traces of a perl interface to that library.17:27
Works for me so far, that I can get the version string out of git and read the ls-tree output in a perl array.17:27
ShadeHawk perhaps pasky would be interested, dst_17:27
pasky ShadeHawk: to keep the conventions17:29
mostly17:29
alp i have a small c# git binding that might be useful, http://www.ndesk.org/tmp/gits.cs17:30
pasky I'm now doing .xs for basically git-hash-object17:30
dst_ I'll put it online. Give me a few minutes to sort it out.17:30
ShadeHawk would you support/do you plan supporting CGI calling conventions, i.e. repository({-directory => "~/git"}); (with anonymous hashref)?17:31
→ GeertB joined17:39
→ anholt_ joined17:39
dst_ My stuff is at http://stosberg.net/pub/gitlib/ . To test, extract the tar inside your git directory and apply the shared-lib.diff. You'll need swig to generate the perl-to-c bindings.17:42
pasky ShadeHawk: what exactly are you after?17:42
dst_: hmm, the create_array() bit looks scary17:43
dst_ Unfortunately swig's calling conventions require this.17:44
It can't hand perl arrays to the c code.17:44
pasky I guess I'll rather stay directly with the .xs then... but the idea of calling builtins thru libgit looks nice17:46
aha, but that's not something linked to libgit by default :)17:46
ShadeHawk to be able to use repository({key => value, key => value, key => value})17:47
pasky ShadeHawk: instead of repository(key => value, key => value, key => value)? what'd be the point?17:48
→ krh joined17:58
pasky damn, this reminds me why I hate git header files18:04
(not to say perl headers aren't orders of magnitude worse)18:06
→ Paavo joined18:07
pasky $ perl -w -I/home/xpasky/lib/perl5/site_perl/5.8.8 -MGit -le 'print Git::hash_object("Makefile", "blob")'18:11
c0f64523c7861ebb6a414c22c58793f0ff35729618:11
woohoo!18:11
dst_ Congratulations.18:13
nickh sweet!18:13
ShadeHawk time?18:14
i.e. $ time perl -w -I/home/xpasky/lib/perl5/site_perl/5.8.8 -MGit -le 'print Git::hash_object("Makefile", "blob")'18:14
swoolley that includes the entire perl process18:15
might want to loop the print and time that18:15
pasky 0.027s vs. 0.32s for generic_oneline18:15
so it's not really impressive but yes, what swoolley said18:15
$ time perl -w -I/home/xpasky/lib/perl5/site_perl/5.8.8 -MGit -le 'print Git::generic_oneline("hash-object", "Makefile") for (1..1000)' >/dev/null18:16
real 0m4.775s18:16
$ time perl -w -I/home/xpasky/lib/perl5/site_perl/5.8.8 -MGit -le 'print Git::hash_object("Makefile") for (1..1000)' >/dev/null18:16
ShadeHawk $ time perl -w -I/home/xpasky/lib/perl5/site_perl/5.8.8 -MGit -le 'print "c0f64523c7861ebb6a414c22c58793f0ff357296\n"'18:16
pasky real 0m0.178s18:16
much more impressive, I'd say ;)18:16
ShadeHawk 25 x faster...18:17
swoolley \n isn't needed with -l18:18
ShadeHawk pasky: is this single hanging semicolon after croak in arg being filehandle case needed? if so, why?18:45
croak("You passed me something weird - a dir glob?");18:45
;18:45
pasky ShadeHawk: it is needed since I'm about to write a command there :)18:47
you are looking at what I'm literally editing right now18:47
ShadeHawk ahhhhh... now I understand... working dir state...18:48
I'm wondering whether to wait for Git.pm before refactoring gitweb18:49
(perhaps someone else would do that ;-))18:49
pasky Git.pm is probably going to evolve fast and gitweb would be a great test-user for all the new interfaces we are gradually going to add18:51
→ z3ro joined18:53
normalperson pasky: looks sweet19:03
→ cworth joined19:19
→ ShadeHawk joined19:21
→ Eludias joined19:22
→ DrNick joined19:41
→ krh joined19:56
pasky so, I've done some pretty big changes :)20:16
to the API and stuff20:16
ShadeHawk: what did you mean by:20:19
(to be converted in the 'new'20:19
constructor I guess)20:20
ShadeHawk my $repo = new Git::repository('/home/jnareb/git/');20:21
or something like that.20:21
of20:22
my $repo = Git::repository->new('/home/jnareb/git');20:22
pasky well, if we'll have an init-db repository constructor, this could get mighty confusing20:26
so I'd rather not20:26
besides, we'd just have to offload everything to a Git::repository object, which doesn't make much sense to me20:26
dormando How does git/cogito handle tracking symlinks?20:26
or could someone point me to the doc file I can't find? :\20:27
→ CIA-14 joined20:28
ShadeHawk dormando: symlinks are supported - see: http://front.fuw.edu.pl/cgi-bin/jnareb/gitweb.cgi?p=git.git;a=tree;hb=HEAD;f=gitweb/test20:30
→ agile joined20:30
ShadeHawk for tree object, symlink is object marked as symlink20:30
for blob, it is just a file name it points too20:31
ate laste that looks from diff: http://front.fuw.edu.pl/cgi-bin/jnareb/gitweb.cgi?p=git.git;a=commitdiff;h=HEAD20:31
diff for creating symlink looks like that:20:32
new file mode 12000020:32
index 0000000..ae3521120:32
--- /dev/null20:32
+++ b/gitweb/test/symlink20:32
@@ -0,0 +1 @@20:32
+file+plus+sign20:32
\ No newline at end of file20:32
dormando Hm. okay.20:34
Had some weirdness just now, but it looks like a cogito thing20:34
cg-rm couldn't remove a bunch of symlinks (I think we're one version behind)20:34
git rm -f did it20:34
ShadeHawk and changing symlink target is for diff:20:35
dormando running through a push/pull cycle to see if it actually got rid of them20:35
ShadeHawk diff --git a/gitweb/test/symlink b/gitweb/test/symlink20:35
index ae35211..f97d055 12000020:35
--- a/gitweb/test/symlink20:35
+++ b/gitweb/test/symlink20:35
@@ -1 +1 @@20:35
-file+plus+sign20:35
\ No newline at end of file20:35
+hardlink20:35
\ No newline at end of file20:35
dormando Looks like it worked fine, sorry :P20:36
ShadeHawk So git should handle symlinks well.20:36
dormando Confirmed that the diffs look like you just described20:36
Seems like cogito has some trouble with symlinks. I'll upgrade it and report if it's still happening20:36
Thanks :)20:37
pasky cogito wasn't tested with symlinks a lot :(20:40
dormando Hm :\ What I tried to remove was 100 symlinks to a directory, and it saw them all as directories20:41
So blah blah.20:41
git rm -f zapped the buggers fine20:41
pasky :))20:42
ShadeHawk pasky: IIRC there was a patch to support --git-dir command line option at least for git wrapper...20:45
in the 3 parts series of making gitweb mod_perl friendly20:45
pasky I'll have a look20:47
ShadeHawk hash_object doesn't work for "tag" type?20:48
do you know if there is a perl module to check if GPG signature matches?20:51
pasky Crypt::GPG as a 10-second CPAN search revealed ;)20:53
hash_object probably works for any object type20:54
not sure about delta20:54
the --git-dir patch just makes git.c setenv(GIT_DIR), so not really useful :(20:54
but extending the setup_repo_whateverisitnamed() function to support that should be easy20:54
chris2 hrm, 1000x makefile hashing is here: 3.059989 (git.rb) vs 71.056134 (git-hash-object) on a totally loaded and slow ibook20:58
ShadeHawk and don't forget to add '--git-dir' support to Git.pm to use by Perl scripts.20:58
chris2: git.rb?20:59
chris2 a ruby binding for libgit20:59
i wrote this afternoon20:59
<100loc ruby, 0loc c. already can lookup sha1, traverse trees and hash objects :-)21:01
ShadeHawk by git-hash-object you mean calling this from inside ruby script, or what?21:02
chris2 git.rb calls index_fd, and git-hash-object forks 1000x times21:03
pasky Temporary fix: commented out offending line in mimetype_guess.21:04
ShadeHawk: what does that mean?21:04
ShadeHawk if yes, this is an order of speedup for Git.pm fork+pipe+exec, vs XS.21:04
pasky: this line caused error (undefined global $path), so I commented it out. I don't know what this line *meant* to do, to correct it correctly21:04
pasky # (relative to the current git repository)21:05
this part of $mimetypes_file21:05
→ spyderous joined21:05
spyderous how can i revert an entire merge?21:06
preferably without individually reverting each commit that was merged over21:06
ShadeHawk 23.2 for your ruby binding, around 26.8 for pasky's solution (only hash_object for now)21:07
chris2 the ibook probably distorts the numbers a lot21:07
ShadeHawk spyderous: I guess git-reset --hard ORIG_HEAD, but I'm not sure.21:07
pasky ShadeHawk: it's different hardware21:07
chris2 if someone has ruby 1.8.4 installed, i can post it21:07
spyderous i already pushed it, so i can't just kill the history21:07
well technically somebody else pushed it, but same point =P21:07
pasky ShadeHawk: even different architecture so relative fork() costs might be quite different21:07
chris2 esp on osx :P21:08
pasky oh21:08
ok, that too :))21:08
chris2 i guess perl xs will be absolutely faster than ruby/dl21:08
but far more work, too ;-)21:08
ShadeHawk chris2: do you plan to publish it/send it on git mailing list...21:10
chris2 it's very ugly atm21:10
ShadeHawk BTW git-svnconvert in Ruby was it your work?21:10
chris2 no21:10
pasky spyderous: decide "which" parent of the commit do you want reverted, then do a diff between the parent and the commit, apply that in reverse and commit21:10
chris2 ShadeHawk: i guess i'll publish it when its a bit more mature21:11
ShadeHawk chris2: if you have some free time, you can compare speedup from pasky's Git.pm encoding hash_object in XS...21:12
chris2 will do later21:12
ShadeHawk pasky: documentation got out of sync with implementation. command does not return filehandle in scalar context21:15
spyderous pasky: ok, thanks for the tip21:15
ShadeHawk IMVVHO generic command() should return array for array, filehandle for scalar, and run quietly for null/void context.21:16
OR return array of lines for array context, first line for scalar context, and run quietly for null/void context.21:17
from the common usages.21:17
if you want to return whole output for scalar context, instead of using array then joining elements, wouldn't it be better to just slutp (undef $/) the file?21:19
→ kampasky joined21:19
ShadeHawk s/slutp/slurp/ (and this is Perl termonology IIRC)21:20
ShadeHawk is confused... it looks like he doesn't understand Perl...21:21
pasky well, this is tough to argue about since this is really matter of taste21:22
but to me, this way it just look much more logical to me - it's kind of how you would expect the function to behave21:23
even if it might not be the most economic way to the number of functions or characters typed when programming21:23
thanks for the documentation alert21:24
ShadeHawk if (!defined wantarray) { local $/; return <$fh>; } or something like that...21:26
pasky huh?21:28
if you don't want anything, the best thing you can do at the point is to close the $fh21:29
ShadeHawk sorry, it should be:21:31
my $content;21:31
pasky is confused to no avail :)21:32
ShadeHawk { local $/; $content = <$fh>; } unless (defined wantarray && !wantarray);21:32
...21:32
return $ccontent; # instead of join('", @lines);21:33
→ cworth joined21:33
ShadeHawk now clear?21:33
s/unless/if/21:34
Gah. Changing code while writing...21:34
i.e. slurp if in scalar context, discard output if in void context, return array of [chomped?] lines if in array/list context21:35
pasky Hmm.21:36
→ kampasky joined21:36
ShadeHawk should be little faster, becaue Perl doesn't need to split output into lines, only to discard it later...21:37
pasky But it will be significantly slower (and O(N) in memory usage) in void context.21:38
ShadeHawk no, slurp only for scalar context, i.e. for (defined wantarray && ! wantarray)21:38
pasky oops21:39
I guess the patch was too huge for the mailing list21:39
hrmz21:39
ShadeHawk { local $/; my $content = <$fh> } if (defined wantarray && !wantarray);21:39
@lines = <$fh> if (defined wantarray && wantarray);21:40
i.e.21:40
@lines = <$fh> if defined wantarray;21:40
or even21:40
my @lines = <$fh> if defined wantarray;21:40
although conditional declaration is frowned upon, if possible at all...21:41
pasky I will just outsource the close to a sperate function and write separate code paths :)21:41
ShadeHawk you added command_pipe, but didn't add it to @EXPORT_OK21:43
pasky oops, another thing I forgot :)21:43
good thing the patch didn't make it to the list21:44
ShadeHawk and you might want to add info about void context for command to documentation, too...21:44
or not, as it is obvious (it is an implementation detail)21:45
...ah those XS names: SVt_RV, SVt_PVGV, IoIFP, sv_2io, SvPV_nolen...21:48
DrNick look at all that wasted Perl effort21:49
you could be doing this in Python, you know21:49
or a selection of many other languages that aren't write-only21:49
ShadeHawk Common Lisp for example? ;-P21:50
chris2 hates the read_tree_recursive call21:51
ShadeHawk DrNick: try to write something in TeX language... that is write-only language...21:51
C for example? ;-))))21:51
pasky ShadeHawk: Yes, when doing XS today, first in a rather long time I felt again as a true wizard. :)21:52
← spyderous left21:52
pasky I just love that about doing XS. Real raw magic.21:53
chris2 apropos magic. will i get slapped for using const char *base in read_tree_recursive to pass a closure? ;-)21:54
pasky eek :)21:58
chris2 that's the most evil ruby i ever wrote22:01
Git.read_tree_recursive(@tree, [block.object_id | (recursive ? 1 : 0)].22:01
pack("L"), 4, 0, nil, READ_TREE)22:01
it places the address of the closure, together with whether to walk recursive in the first 4 bytes of the base ;)22:01
pasky cute :)22:03
chris2 actually, there is a worse line atm22:03
pasky hmm, I should check ruby out.. the syntax seems quite similar to perl22:03
chris2 rstat = File.stat thing22:04
stat = DL::PtrData.new(rstat.object_id*2).to_a('P', 5).last22:04
this calculates where the struct stat* is placed internally ;-)22:04
pasky or perhaps just the @tree part appealed to me ;)22:04
chris2 you'd probably enjoy it22:04
@foo means instance variable22:04
there are no sigils for datatypes22:05
only for scope22:05
ShadeHawk pasky: Patch (Git.pm v3) passed22:07
dormando okay... symlink annoyance again22:08
pasky ShadeHawk: passed? :)22:08
dormando we have remote repo origin22:08
user A git rm -f's a ton of symlinks22:08
user B cg-update's22:08
it doesn't remove the symlinks22:08
user B commits and pushes, and the symlinks go back into the repo22:08
pasky that's weird22:09
dormando One of the devs is about to put my head through the floor :P22:09
pasky did the cg-update do a merge?22:09
dormando sec22:09
possibly. When I do a fast-forward the symlinks poof22:10
lemme go bug user B22:10
pasky it's not that important as what cg-status shows after the cg-update22:11
ShadeHawk pasky: Hmmm... if you use git wrapper, using "--git-dir=$self->{opts}->{Repository}" as a first arguments can be alternative to setting $ENV{GIT_DIR}22:14
i.e. is on git mailing list22:14
pasky ah.. cool22:14
ShadeHawk: well, I kind of prefer the $ENV{GIT_DIR} solution since it doens't clutter ps axu output ;)22:14
it might be a wee bit slower but I think that's not gonna be noticeable in all the fork() overhead22:15
dormando pasky; ok, sec...22:15
bugging people22:15
I think it just doesn't show them as changed at all22:16
ShadeHawk that is supposedly needed for mod_perl compliant gitweb... it could be set as an option when creating repository... ergh, perhaps one should separate repository from git configuration.22:16
dormando I see he did a merge earlier, but most are fast forwards22:16
and saw nothing in his backscroll about the symlinks22:16
pasky ShadeHawk: well, I claim it isn't22:16
ShadeHawk: remember that we set the %ENV already in the subprocess22:17
ShadeHawk ah... yes...22:18
By the way, for "git-command" invocation one needs gitexecdir (better to remember it, instead of calling git --exec-path every time)22:18
pasky yes, gitexecdir is what I want to add now22:18
ShadeHawk Git.pm is very nice.22:22
By the way, are 'git' always in gitbin aka bindir, _and_ in gitexecdir aka libexecdir?22:23
pasky I don't think so22:26
Git::exec_path() implemented22:27
dormando pasky; this is ... really annoying :(22:35
Any suggestions on how we should fix the repo?22:35
we need to get rid of the symlinks, but anyone else doing work adds them back in22:36
→ tmokros joined22:39
ShadeHawk pasky: could you also add Git::git_version()? The only cases where we might want to call 'git' directly, and not one of the git-commands.22:43
pasky ShadeHawk: sure22:44
ShadeHawk attention: it should be git version, as we might want tu use script for testing from never version that the one installed.22:44
that's the beauty of scripts.22:44
pasky dormando: I have to admit that I didn't ever track symlinks with Git and I'm afrait to lose my mental state now so it might be best to ask around on the mailing list22:45
ShadeHawk I use git 1.3.0 (Slackware ready package from LinuxPackages), but git-send-email from current version, and muck with current gitweb.cgi22:45
so relying on @@VERSION@@ might be out (but might be first reslt if set)22:46
pasky ShadeHawk: actually, I'm ending up abolishing GitPath again and will not make use of exec_dir(), and rather call any external commands using execv_git_cmd() ;)22:46
ShadeHawk: well, if your .pm is out-of-sync, that means most likely .so of your .xs is out-of-sync, and that means everything will behave _really_ strange, especially as random functionality may shift between external commands and the .so22:46
ShadeHawk: so, the .pm you use must match the git version you have installed22:47
dormando pasky: no problem22:47
ShadeHawk Oh well...22:49
well, you might not make use of git_exec_dir, but you might leave it and git_version, even if unused.22:53
pasky sure, I won't retract the patch22:53
ShadeHawk because for example at least for now Git.pm doesn support: command_in, command_inout, and creating pipelines (although you might do it by hand-stitching I guess)22:55
pasky the trouble with XS is that my fingers can't decide whether to write continue or next22:56
→ yacc joined23:00
yacc Just wondering, is there a git port to Win32?23:00
→ idr joined23:12
idr I've encountered an odd situation with git, and I can't figure out how to make forward progress.23:13
I have commits in my local tree to a file. That file was subsequently deleted upstread, and I did a 'git pull'. File gone.23:13
Later, the file was re-added (or the deletion was undone?) upstream.23:14
Now 'git pull' has problems on that file.23:14
Specifically, I get foo.c~HEAD and foo.c~[commit id gook] where foo.c should be, and git tells me "Automatic merge failed; fix conflicts and the commit the result."23:15
anholt_ so, you move foo.c~HEAD over fooc, git-update-index foo.c, commit, right?23:16
idr If I try to commit back a version of foo.c that is what I want, git tells me "Cannot do partial commit durning merge."23:16
anholt_: I did try that also.23:17
anholt_ what did that do?23:17
idr anholt_: But, IIRC, the commit said no changes.23:17
anholt_: Let me try it again...23:18
anholt_ what exact commit command did you use?23:18
idr Okay.23:18
anholt_: Doing 'git pull', edit of xf86sym.c, git-update-index xf86sym.c, git commit xf86sym.c results in ""Cannot do partial commit durning merge."23:19
anholt_ just git commit23:19
you want to commit the whole merge, not your update to a single file.23:19
idr anholt_: I guess that makes sense. And it did work, too.23:21
anholt_ (and git-update-index is saying "I've fixed the conflicts in this file, it's ready to go". git status will tell you if you have further files to fix up)23:21
→ Simon- joined23:21
← Elrond left23:25
← Simon- left23:26
→ benlau joined23:32
→ mchehab joined23:51
yacc Just wondering, is git available for Windows?23:56
neuralis yacc: should work fine under cygwin, don't know about native binaries23:58

Logs Search ←Prev date Next date→ Channels Documentation