[TYPO3-dev] The future of Bug Days

[Jigal van Hemert]

Hi,

On 25-7-2011 8:10, Georg Ringer wrote:
> Am 22.07.2011 15:10, schrieb JoH asenau:
>> Just a wild guess, but could it be that the decrease of particpiants is
>> not due to routine, but just the contrary: Missing routine with the new
>> git/gerrit-concept, which is at least overly complicated to set up
>> compared to the "old school" SVN, people are used to?
>
> but on the other hand it is more used to new people who got their
> projects e.g. at github.

My guess is that the percentage of new contributors who have a project 
on git(hub) is very small.

> Of course I agree that it is more complex on the other hand but now also
> far easier to test a review. This was always *very* hard in old days
> when there were conflicts (which was there most of the time when patches
> where a bit older)

Ehmm... my experience is completely the opposite. At least the way 
TortoiseSVN handled the "Apply patch" option was very flexible. Most of 
the times it could easily find the code it had to change, even if it had 
moved a substantial number of lines in the file.
A git "change set" is a snapshot of the entire branch and thus is is 
much more likely to produce conflicts (e.g. when submodules are changed).
In case of conflicts TortoiseSVN has a wonderful tree-way diff tool to 
solve conflicts (where you can decide to keep lines from patch or 
current version and/or edit the lines)

Reviewing a patch with svn was for me as simple as:
- Apply patch
- test
- Revert
If I was the dev to commit the last step would be replaced by:
- edit changelog
- commit

> I am using windows too and using command line is IMO the best and was
> very easy to set up.

If you are not a command line person you are basically left on your own.

> i loved the mailinglists but it was also very hard to have an overview:
> what got enough votes, where are new patchsets and you also could never
> compare diffs which each other.

A decent news reader such as Thunderbird (but others will probably have 
comparable features) has the option to attach categories (tags) to 
messages or threads. With a single click you can filter these. In a 
thread with the tag ("reviewed by reading and testing") a new message 
would emerge with another positive vote and I would simply mark it 
("need to commit").
Okay, not technically advanced, automated, etc., but it worked very 
well. I could easily find patches I needed to review or commit.

In Gerrit I find it very hard to find change set which I have or have 
not reviewed yet, especially when a new version was pushed.

> revitalize for discussing would be nice. so for each review a new thread
> starts but IMO nobodoy got the extra time to get patches svn style into
> gerrit.

And if it would mean that reviewing activity would increase a lot? Would 
that justify the extra time?

> I don't know if bug days have something to do with git/gerrit?

Activity in both bug days and development has decrease a lot since the 
introduction of git/gerrit. Causation is hard to prove, but it is so far 
the only explanation which can be found...

-- 
Kind regards / met vriendelijke groet,

Jigal van Hemert.