[Kde-scm-interest] initial port of scripty to git is ready

[Matt Rogers]

On May 21, 2009, at 2:37 PM, Chani wrote:

> On May 21, 2009 06:39:29 Matt Rogers wrote:
>> On May 21, 2009, at 8:16 AM, Ian Monroe wrote:
>>> On Wed, May 20, 2009 at 10:34 PM, Matt Rogers <mattr at kde.org> wrote:
>>>> Hi,
>>>>
>>>> I've finished an initial port of scripty to git. It can found on
>>>> Gitorious at http://gitorious.org/scripty. There are several  
>>>> issues,
>>>> if you want to call them that, with it:
>>>>
>>>> 1. completely untested
>>>
>>> pinotree says they only way they test is very carefully with the  
>>> real
>>> scripty. This makes me sad. Maybe we could get an anonsvn server
>>> cloned and made writable, then we could have something to test with.
>>>
>>>> 2. there is no initial git clone to get a working copy if it is
>>>> missing since I don't know what the layout of the new git
>>>> repositories
>>>> is going to look like.
>>>> 3. There is no 'git push' back into the main repository from
>>>> scripty's
>>>> clone
>>>
>>> So that we're on the same page... scripty still needs to work with
>>> i18n in SVN. I was a little concerned when you said its a "git port"
>>> when its really more of a "git feature addition." I guess I  
>>> should've
>>> dented immediately. :)
>>>
>>> The idea is that for Amarok or Plasma to switch to git tomorrow,  
>>> i18n
>>> needs to stay in SVN. (Possibly i18n could stay in SVN forever, they
>>> have a very different way of doing things and switching to git  
>>> should
>>> be up to them.)
>>>
>>> The only thing that needs to be git pushed are the .desktop files  
>>> that
>>> scripty famously commits in our SVN repos routinely at the moment.
>>>
>>> Ian
>>
>> *sigh* brilliant. too bad nobody told me this initially when i said i
>> would do it.
>
> you never asked! ;)
> this is why I'd spent several days poking at things and thinking and  
> asking
> questions instead of getting work done. unfortunately I never wrote  
> down what
> I learnt, and I was away all yesterday while you were apparently  
> doing the
> work. faiil. :)
>

True. I'm happy to pass the work on to somebody else. However, I told  
Ian that I would work on it, and the only thing I remember from when  
he and I talked about it back in March is "port scripty to git" so  
that's what I did. And I also didn't understand what Albert was trying  
to say to me yesterday.


> my general impression of things:
> -kde isn't going to move all at once, and there's no reason to force  
> the
> translators to switch to git before they're ready, so it makes sense  
> to leave
> scripty in svn for now, and give it the ability to pull multiple git  
> repos
> (remember, kde's not going to be in one giant repo, git doesn't work  
> that
> way).
> -there are some things, like the svn "add it if it's not added"  
> stuff, that'll
> be easier in git... actually, I bet it could be simplified for svn  
> too, because
> there's no harm in adding a file that's already there.
> -checking whether we're in a git or svn repo before each vcs  
> operation is
> gonna get annoying, so there'll probably be a new script to handle  
> that.
> -it'll be easier to leave the existing scary-code untouched for the  
> most part
> than to rewrite the whole thing. especially as we don't want the svn  
> bits to
> break. keep the status quo as much as possible, then look at  
> improving things
> once people are comfortable with the move-to-git step.
> -I'm assuming scripty will run on a machine with a checkout of all  
> kde code
> -I looked into other existing translation systems, but they're web,  
> and our
> translators don't want web interfaces, so that settles that.
> -Chusslove has told me about gnome's decentralized way of doing  
> translations,
> and it sounds like a lot more work for project mainatiners. I think  
> scripty is
> our best option. centralized translation seems to be an easier way  
> for both
> developers and translators.
>
> skimming through some of your code, I think I found a bug:
> check_desktop_files:210:
> system("git update-index --chmod=-x -q propdel '$filename'");
>
> that propdel shouldn't be there, right? :)
>

nope. porting error.

>
> btw, I don't see why people want to test this on amarok *first*. why  
> not make
> an itty-bitty dummy project on gitorious, put some dummy cpp and  
> desktop files
> in it (possibly just copy a few from a random project and change the  
> strings
> to have SCRIPTY_TEST in them), and test on that? :)
>
>