[Geany-devel] Git conversion completed

[Colomban Wendling]

Le 09/10/2011 19:03, Jiří Techet a écrit :
> On Sun, Oct 9, 2011 at 17:59, Colomban Wendling
> <lists.ban at herbesfolles.org> wrote:
>> Le 09/10/2011 16:49, Jiří Techet a écrit :
>>> On Sun, Oct 9, 2011 at 15:38, Colomban Wendling
>>> <lists.ban at herbesfolles.org> wrote:
>>>> Le 09/10/2011 14:36, Jiří Techet a écrit :
>>>>> On Sun, Oct 9, 2011 at 13:55, Colomban Wendling
>>>>> <lists.ban at herbesfolles.org> wrote:
>>>>> [...]
>>>>>> * 03c3b75 (r3679) is a bit weird too, it has d3cdd27 (r3680) as parent,
>>>>>> but as the SVN revision suggests, d3cdd27 is newer than it (13:53:04 vs
>>>>>> 13:45:47).  And I don't see from SVN log why this would be wanted.
>>>>>
>>>>> This is apparently my error - judging from the date I should have used
>>>>> d71d352 for "merge trunk changes". I'll fix that.
>>>>
>>>> Great, looking forward to it.  I think when this is done all is OK, so
>>>> we could import it in the official repo :)
>>>
>>> OK, I've re-uploaded the repositories to the same locations. I've
>>> fixed the incorrect merge issue and updated parents for most of the
>>> re-created branches.
>>
>> Yep, looks great, bravo!
>>
>>> The only one I haven't updated is the "Create branch for configurable
>>> build menu development", now commit 80d2802. There's something strange
>>> - it appears the branch existed before but it was never merged into
>>> trunk. Instead it was probably deleted and re-created again. If I
>>> updated the parent to be the trunk only, we'd lose the history of this
>>> branch because we couldn't get to the previous commits in any way. So
>>> I think it's better to keep it the way it is.
>>
>> OK, makes sense, even though the old build-system branch was dropped
>> because it was "corrupted" (according to r3939: "Removing corrupted
>> branch").
> 
> Ah, OK, I've overlooked this. If you want to make some more changes,
> create a file called "grafts" inside .git/info. Each line of this file
> has format
> 
> parent child1 child2 child3... (using commit SHAs depending on number
> of children)

Isn't it rather child parent1 parent2...?  Looks like it does a better
job, if I understand the result correctly ^^  anyway, thanks for the tips :)

Anyway I chosen to keep the previous branch as you did it so there is
some more history, yet it is strange.

BTW, pushed to https://github.com/geany/geany :) -- this is not yet a
wide announcement because there is a few thing to tune, update & qtuff,
but it's up and running.

> When you're finished with your modifications, run
> 
> git filter-branch --tag-name-filter cat -- --all
> 
> to write it permanently into the repository. Remove the grafts file afterwards.
> 
>>
>> I also checked the repository against my own git-svn clone as you
>> suggested, and all about 60 checkpoints in the history I checked did
>> match, so I'm happy to see you didn't corrupt the sources ;)
> 
> And the remaining 5000 commits in between contain code which will
> erase your hard drive completely ;-).

Yep, and you're a so great mind-reader you you knew very well which
commits I would check :p


Cheers,
Colomban