[TYPO3-dev] I blame Git [was:Slow BE since 4.5.x]

[Christian Kuhn]

Hey,

On 04/25/2011 04:00 PM, Patrick Gaumond wrote:
> On 2011-04-21 16:45, Steffen Kamper wrote:
>> i complained a lot about too less action from the devs, and the action
>> doesn't raise at the moment, i have the feeling it lowers more ...
>
> I would blame Git at the moment. And no, there's no smiley to my remark.
>
> Git may be more efficient from a CoreDev point of view but it also
> killed an essential part of CoreList spirit... Discussions and
> interactions. Even if in theory it was prohibited, the lack of "human
> factor" cut out the fun.

well ... human factor has two sides, for some 'old' core patches it did 
not necessarily make me more happy to have a discussion of 50 mails for 
one issue. Discussions are not always focused, they tend to diverge. The 
new review system helps to stay concentrated on ad diff. I still have a 
simple patch at hand that _must_ be fixed, but never ever gets reviews 
after 30 Mails in the old system, numerous +1, numerous -1, about 10 
reminders, even the solution is still the same.

I just pushed https://review.typo3.org/#change,1722 to gerrit. This one 
took me about 20 hours, including 3 pre-patches to typo3 v4, I had to do 
before this darn file backend from FLOW3 was compatible with TYPO3 v4 
again. I've already merged 2 of them (they were splitted in a way that 
they could pretty much go through as no-brainers after a first positive 
review by a sane person), the third one and the main patch are pending. 
I hope they will get reviews soon, since all my next patches must be 
based on this. Next to this I pushed 5 patches back to FLOW3, pretty 
much all of them can be considered as no-brainers. They took me nearly 
zero time (ok, the unit test mocks in one of them didn't not work out as 
quick as I have thought).
But it will stay this way: All my next patches will be a chained, based 
on each other, to finally reach some goal. I need to get used to handle 
this 'the git way', but with svn, those 'create-on-top-of' patches are 
simply impossible.

Next to this, it was easily possible for me to simultaneously keep track 
of other issues by different persons, push new versions, comment on 
issues and maybe finally merge them. This was exactly no problem with 
git/gerrit for me after same training, and I probably would not have 
reviewed those patches without this system.

So, at the moment, git/gerrit currently helps me more than it sucks. I 
hope we will get more reviewers with a feeling of good code, who push 
updated versions and review things, though.

We will see if we benefit from this in the long run.

My 2 cents
Christian