Re: gerrit for 3-6, LibreOffice gerrit bot & wrong Merged status

On 15.07.2012 12:29, Bjoern Michaelsen wrote:
> On Sat, Jul 14, 2012 at 08:02:37PM +0400, Ivan Timofeev wrote:
> > >I am trying to use the following workflow on master:
> > >
> > >1. git commit
> > >2. ./logerrit submit libreoffice-3-6
just fine till now.
> > >3. git push
Huh?! What is that for? it makes no sense for me to simultaneously 
submit to gerrit and to push directly to master (and thus bypass gerrit).
Think about it: if your patch is going to be rejected or must be 
adjusted what is going on with your changes that already in master?
Note 1: we are going to turn on the automatic patch verification 
machinery (build + test). What is if your patch doesn't even compile, 
introduces WaE, beaks unit tests?
Note 2: in the long term (unsure how long, though ;-) there will be no 
way to push directly to master (and bypass gerrit) anyway.
> Since the gerrit team seems to be struck in the philosophical
> question if a change is still the same on a different branch, I guess we should
> patch our gerrit(*).
I took a look into that patch: it closes the (first) gerrit patch only 
if in the submitted (second) master commit the branch information is the 
same.
But, as you pointed out gerrit can not handle it currently anyway (*), 
why not to follow this simple workflow:

*Either* to push the patch directly to the master and send a mail to ML 
with cgit link and ask for review -- my i add not recommended here? ;-) --,
*or* to submit it only to gerrit (master or specific branch), wait for 
verification and once approved and merged,
push it to all other branches (manually for now or even through gerrit 
later).

Regards
David

(*) I disagree with you to patch our gerrit because of that.