On Tue, Sep 14, 2010 at 9:13 PM, Carsten Haitzler <raster@...> wrote:
> On Wed, 15 Sep 2010 09:05:15 +0900 Mike McCormack <mj.mccormack@...>
>> On 09/15/2010 06:49 AM, Carsten Haitzler (The Rasterman) wrote:
>> > On Tue, 14 Sep 2010 21:18:58 +0900 Mike McCormack<mj.mccormack@...>
>> > said:
>> >> On 09/13/2010 08:46 PM, Gustavo Sverzut Barbieri wrote:
>> >>> Yeah... that's the kind of problem I said in the other mail thread
>> >>> that got disturbed by stupid GIT discussions :-/ See guys, not even
>> >>> GIT would help here :-/
>> >> The problem here isn't the revision control system, it's that
>> >> there's no systematic code review before code goes into SVN.
>> > if you want development to work at 1/4 the speed it does... then just keep
>> > suggesting that. :)
>> I don't really buy that argument, because breaks keep developers away
>> and slow
>> down development anyway, but it's your project...
> i do - because i actually do have some level of patches that end up in queues
> (for people without svn access) and they eat up quite a bit of time. even for
> the few that are around and even for fairly superficial review.
>> > review doesn't just magically catch all bugs.
>> It magically catches 80% of bugs, which is a good start.
>> How about having patches at least verified by a buildbot before they're
> that's the job of the committer to at least have built the src first and run
> and tested it. if devs consistently commit patches that dont even build or work
> in the most basic way then they probably should have svn access revoked. this
> isnt a technical issue - it's a human issue. to be solved the human way - spank
> them a few times and eventually kick them out.
I like this idea... We're doing such stuff for webkit-efl based on
contract with Samsung, maybe someone at Samsung could implement (or
have us to) something similar in EFL?
The webkit way is also using svn, but instead of directly commit
people have to create bugs, then the bot goes and compile the patches,
if it fails then a highlight shows that. If not then people add a cq+
in bugzilla and it will be magically merged into svn (if it compiles,
if something else introduced a conflict it will be rejected and
Of course they have scripts to do so and also to force
unreviewed/unchecked fixes, like we are doing to recover from their
EFL build breaks (as we do not have the bots connected to their system
Sounds a big thing, but if we can have someone to do it, then it will
not impact the developers a lot.
Gustavo Sverzut Barbieri
http://profusion.mobi embedded systems
Mobile: +55 (19) 9225-2202