IRCloggy #git 2006-04-17

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.

2006-04-17

→ Pug|Lagged joined01:07
→ pasky joined01:32
→ Pug|Lagged joined01:45
Pug|Lagged → PugMajere01:45
→ benlau joined02:29
→ loops joined03:47
→ cworth joined03:50
→ Oejet joined04:40
→ hharrison joined06:22
→ ferdy joined07:16
→ juvenis- joined07:40
→ ferdy joined08:55
→ hharrison joined10:26
→ morans joined11:30
→ coywolf joined11:47
→ Aexoden joined11:57
yann pasky: just tried cg-admin-rewritehist to make a graft permanent, and get (cg 0.17.2, git 1.2.6):12:31
$ cg-admin-rewritehist -r master neworigin12:31
head: error reading `../map/': Is a directory12:31
usage: git-update-ref <refname> <value> [<oldval>]12:31
cat: ../map/: Is a directory12:31
Rewritten history saved to the neworigin branch12:31
(and there is no "neworigin" branch afterwards)12:31
pasky uh12:52
please try latest cogito git tree12:52
and there are some outstanding bugs anyway, I think12:53
← coywolf left13:12
← Oejet left13:21
→ boto joined13:39
spuk-_ → spuk-13:43
yann pasky: cogito-0.17.2-ga433433 with git-1.2.6 gives the same result14:01
oh, a433433 is not official, it is master with my "accents" patch14:02
pasky I'll have a look in 0.5 hour14:04
yann pasky: currently investigating14:05
any reason you don't "set -e" your bash scripts ?14:06
pasky because I didn't at the very start and now it'd be some work to review the scripts for -e safety14:07
and I keep forgetting to do it at least for new scripts :)14:07
yann the problem here is that the "revs" file get created empty, and that case is not handled14:08
pasky because it should be empty unless you are doing -r HEAD14:09
yann it may be normal that "git-rev-list --topo-order HEAD '^master' master" returns nothing, after all14:09
pasky huh14:09
how did _that_ get there, anyway?14:09
yann hey, it's *your* script :)14:09
pasky hmm14:10
it seems like I really wrote it there14:10
any idea what was I thinking? ;)14:10
oh14:11
yann maybe you meant ^${OPTARG}~1 ? (just a wild thought without any particular reason)14:11
pasky yes14:13
^$OPTARG^14:13
:)14:13
CIA-14 Cogito: pasky * r396abf531436 /cg-admin-rewritehist: Fix cg-admin-rewritehist -r14:14
pasky ta14:14
→ timlarson_ joined14:14
dwmw2_gone do we have a way to clone a remote tree but use a local 'alternates' directory?14:14
yann that's called a "japanese smiley patch" :)14:14
dwmw2_gone other than cloning the local tree with git-clone -s, then making a new branch or something for the remote and pulling that, then making it the 'master'14:15
yann pasky: still some problems:14:17
$ PATH=../localgit/bin:$PATH cg-admin-rewritehist -r master neworigin14:17
8fd492e612d4f55b555dc94b6615ef41b0bf1dac (1/1) cat: ../map/05ea4d3b32ca9d57a0d54f05ee0724e6af8459bd: No such file or directory14:17
Committing initial tree e50d4d6d3904b32e9c50e238e4eed145726bfcac14:17
ff9fdb394579c81537dd604ed78c48232e85619e14:17
Rewritten history saved to the neworigin branch14:17
... and the new "neworigin" head points to an empty orphan commit with message of the "master" commit14:19
but hey, there again set -e would stop things before the commit :)14:20
→ datrus joined14:23
← datrus left14:23
→ datrus joined14:23
pasky ah14:24
someone sent me a patch for this14:25
my patchqueue is non-empty14:25
will go through it asap14:25
yann pasky: oh, you're right about set-e-safeness - what makes this run stop is pick_id14:26
never use "[...] && do_that", always "[!...] || do_that" ...14:27
→ alec joined14:28
← alec left14:28
→ benlau joined14:36
dwmw2_gone hm.14:37
dwmw2_gone finds a tree containing14:37
dwmw2_gone Signed-off-by: J�rn Engel [email@hidden.address]14:37
we really ought to reject non-UTF8 input in commit messages14:38
yann pasky: my set-e patch may need closer examination - it would be easy to introduce an error when doing such conversions14:48
pasky hmm, I'm not sure I like it15:03
I find the original form more readable15:03
yann right, in many cases I find the equivalent if construct more readable - would that qualify ?15:23
pasky in some cases yes, but not in all15:23
yann I tend to keep || as an idiom for "assert", thoguh15:23
though15:23
pasky ponders15:24
yann things like "[ assertion ] || die"15:24
that's the only case where I still use them myself, since I find them still readable, and shorter than if15:25
pasky later in the evening I'll go through the patch queue15:25
after a bit of thought, I think I will accept most if not all the changes15:25
yann you mean, those in my "set -e" patch ?15:27
pasky yes15:28
→ YGingras joined15:29
yann cool, less work for me :)16:00
← benlau left16:13
→ dst_ joined16:53
→ tcsager joined17:02
→ dst_ joined17:19
← tcsager left17:19
→ acb joined17:41
→ anholt_ joined17:50
→ schofield joined18:48
→ juvenis-- joined19:09
→ YGingras_ joined19:15
YGingras_ should I use raw git, cogito or qgit for simple everyday hacking ?19:16
YGingras_ is a total noob19:16
loops qgit is just a visualization tool19:16
so the choice is really between git and cogito19:16
cogito needs git installed to run19:17
so you can install git and see how it goes...19:18
if it starts to bend your brain, install cogito and see if it makes more sense19:18
YGingras_ loops, that makes sence. This might sound trollish but how does git compares with bzr? My little experience shows that bzr is extremly slow and git pretty fast. Is there a drawback for all this speed ?19:19
dormando you get your work done faster.19:20
then they fire you when there's nothing left to do19:20
:|19:20
loops i don't know bzr, but the drawback of git might be a steeper learning curve19:20
although in the end, git is really rather straight forward19:20
just a bit different than many peeps are used to19:21
→ mchehab joined19:25
yann loops: no, qgit is not *only* that, it is wrongly listed as such, that's all19:45
YGingras: and you also have the option to use stgit19:46
In most cases I use stgit and a couple of cogito commands19:46
btw, qgit has support for basic stgit operations, but you'll need latest rc version of qgit to use stgit 0.919:47
YGingras_ somehow I don't like the idea of using pre-beta programs when they are responsible of my code storage and integrity ; )19:48
→ robfitz joined19:49
yann YGingras: it is git being responsible of that, qgit only runs git/stgit commands19:50
and "rc" is a"release candidate", not a "pre-beta" :)19:50
only on the linux kernel have I seen "rc" used for "pre-beta"...19:50
YGingras_ : )19:51
yann anyway, with cogito and stgit you already have much things to discover19:52
my advice would be to get familiar with those first. To use core git tools, you'll most likely have to use more non-standard concepts (use of the index, etc), so you can leave that for later19:53
YGingras_ In fact I'll probably end up doing mostly centralized dev but I like to keep the decentratlized option I in the end its much easier for contributors19:56
loops stgit isn't really what most developers would be looking for centralized dev19:57
it's mostly about patch management against a central repository that you don't control yourself19:57
if you manage your own repo, not much reason for a patch stack19:57
yann oh yes there are reasons!19:58
loops and with a modern version of git, you don't really have to think about the index etc19:58
git commit <filenames>19:58
no need to update index etc..19:58
yann, oh i'm sure there are some corner cases..19:58
yann I use it mostly to be able to work on several things at once, and reorganize things in small patches19:59
loops git can do that pretty easily as well, use multiple "topic" branches and then git-cherry-pick19:59
yann afterwards you can just "stg commit" to your centralized branch - that's even a command you would not want to use in the case you mention20:00
loops don't really need patches20:00
yann loops: I bet you haven't used stgit enough, or you wouldn't be doing that comparison :)20:00
loops yann, i don't think you're talking about the use case that YGingras is asking about20:01
yann and about "git commit", yes, there is the porcelain-layer-inside-the-core-package which I do find confusing as hell20:01
loops yann, once you use it a bit, it's easy20:01
i'm not knocking stgit, just saying its not for every use case20:02
yann sure, it does not do everything, that's why I also use cogito - but once you got used to stgit, not having the stack management stuff at hand is frustrating20:03
and using topic branches is no comparison - how would you get the output of "stg series -d" using topic branches ?20:03
loops why would you want to output a stg series -d ?20:04
just merge the topic branch with your main branch when its ready20:04
yann we're just not talking about the same thing20:04
loops you're right, you're talking about sending patches into a maintainer, and i'm talking about maintaining your own repo20:05
yann not at all20:05
:)20:05
loops lol.. well educate me then, why do you output a stg seiies -d ?20:05
series*20:05
yann I maintain my own repo at work (a git mirror of a cvs tree), and use a distinct working tree where I use stgit20:06
loops yeah, the git importing of cvs tree's is heaven20:06
yann my job is build manager - I have to work on several aspects of the build system, but want all my changes in the same tree20:07
(no it isn't, it's deadly buggy, but I know that and can live with that)20:07
loops hmm, works for me20:07
yann so I have a stack of distinct to-be-cvs-commits, and add to them incrementally until I'm happy with them (builds, tests pass, etc)20:08
loops yes, see you're exporting patches to cvs.20:08
an external repo20:08
so there's no doubt stgit is the right tool for the job20:08
yann then I do not use "stg commit" but cvs-exportcommit because the backend is cvs, but the idea is the same20:08
loops but if everything is being maintained within git, there's a lot less reason for patch stacks20:09
yann not really external - I could do the same job with only cvs (just much less efficiently)20:09
loops yes, i just meant.. external to the git repo20:09
yann I use a second repo just as a working area20:09
it's just a way of using the tools, nothing to do with the high-level workflow that my manager can see20:10
loops well, when i make a commit to a git repo, i'm done with it, i don't have to rearrange it later.20:11
just different work flows20:11
yann and about git-cvsimport: I customarily see (or rather do not see) patches in cvs that it does not import back20:11
loops really? that's worrying20:11
yann I guess it has to do with considering timestamps to decide whether a changeset has been imported already20:12
loops ah, so it's more to do with incremental updating.. mostly i've just used it to suck in a cvs repo and then retire the repo.20:12
yann and I suspect the problem to come from the fact I work over nfs20:12
right20:12
loops so, qgit has some stgit integration now ?20:13
looks like the new version is worth checking out20:16
→ hharrison joined20:25
yann loops: right - still only has basic functionnality, but makes stgit available to people allergic to command-line :)20:30
→ DrNick joined20:59
YGingras_ I can't seem to build the doc on Sarge : \21:02
xhtml11 seems to bork21:03
I'll try with only man21:04
noop: "Document /tmp/git-1.2.6/Documentation/git-fetch.xml does not validate"21:05
can I use qgit to pull changes from a remote git archive ? (I don't know if archiche is the right term)21:10
ok I don't if this is even a git usage patern. Say dev A has a working branch, dev B clones it and hacks a lot then he have something working and tags it. He then informs dev A that he can merge all the stuff up to that tag. Does it makes sense ? Does B have to make a serie or patches with format-patch ? Can all pull the patches himself ?21:17
Sorry if this is documented elsewhere21:17
ok, I got it: cg-fetch21:23
← mchehab left22:07
→ spuk joined22:43
mugwump YGingras_: yes, but then dev A will need to merge dev B's branch in subsequently22:46
→ spuk_ joined22:50
YGingras_ ok, now supose I borked my master head, using cogito how can I get back to the last working commit ?22:55
well, I see how I can get back using cg-seek but how can I make this commit my new master head ?22:59
→ somegeek_ joined23:03
pasky cg-switch -f -r COMMITID master23:04
→ lilo_ joined23:04
pasky it's mentioned in cg-seek documentation23:05
YGingras_ oh, --long-help ...23:05
→ spuk joined23:06
YGingras_ pasky, does "cg-switch: switch blocked: seeked from master (some commands can be still forced)" mean that I'm screwed ?23:07
pasky just cancel the seek23:07
YGingras_ ok, a plain "cg-seek" than "cg-seek -f -r ..." ?23:08
that seems to wokr23:09
woohoo! cg-diff -c is neet! : D23:11
why does cg-merge autocommit? don't you usually review changes before you commit a merge %23:12
?23:12
pasky usually I review them before cg-merge :)23:14
cg-merge -c will disable autocommitting23:14
hmm, that's a silly flag letter23:14
I'll rename it to -n23:14
CIA-14 Cogito: pasky * rc21f003d0ae6 /cg-merge: Rename cg-merge -c to more consistent cg-merge -n23:16
YGingras_ pasky, support I have a file that I merged and there is a hunk that I want to merge and one that I wand to reject, is there a convenient way to edit the diff? qgit+konpare doesn't seem to affect the working tree23:17
pasky, actually I would have like to make that change just to get the feel for a real life branch+commit23:19
pasky you can edit at the commit time by doing cg-commit --review23:20
but commit time is quite late23:20
earlier, not yet... but it's planed (as "cg-shelve" or something)23:21
+n23:21
YGingras_ that would be great23:21
I can get the file that I want with qgit+kompare but the save button won't save it the working tree, I need to do save as. Kompare is not that good, xxdiff is easier to understand though uglier23:23
pasky, for now how should I refuse a merge? "cg-diff > foo.diff && emacs foo.diff" than reverse patch the unwanted changes ?23:30
I mean partially refuse a merge23:30
pasky yes, probably23:33
YGingras_ ok, I think I'm up to speed. git+cogito is great. : D23:34
pasky, how do you usually develop? You make a devel head and when it passes integration tests you move it to master or you just tag master when you have a solid integration ?23:36
pasky I personally just develop on master ;)23:39
YGingras_ pasky, its pretty impressive that you manage to do all of this in shell script but how will you handle i18n?23:44
I would actually be interested in doing the french translation23:44
→ Euler joined23:50
pasky there's a gettext tool, and in the long (possibly very long) term we will probably move to perl23:54
YGingras_ sounds good23:58

Logs Search ←Prev date Next date→ Channels Documentation