Hi all,
just my first impressions about gerrit, hoping it could be useful
to improve the developer experience.
* Registration with OpenID: no all providers are accepted: openid.org
(first choice, seen suggested on a mail for initial announce), was
not. It could be useful to add a list of supported providers in the
instructions.
* Single patch submission: went smooth, no problem at all.
* Set my rep for further work: it took me some time to understand that
I had to execute a "nextchange" command, I initially thought it was
necessary to sequentially submit successive patches if more than one
was queued. I had to give a look to the logerrit script to clarify
what it was needed for.
* Fixes for the submitted patch. Not so easy to do, after a nextchange
(but I could have been over-complicating the process). From the
"nextchange" study I understood that backups were created, with some
internet-search I got to the commands that restored the backup branch
as my master, then I applied the required modifications and finally
pushed the new revision. The last part of the process was quite easy,
but I would say that some logerrit commands to restore a backup and
commit the patch would be useful. Something alike:
* logerrit backups - to list the available backups
* logerrit restore <desired backup> - to set the desired backup as
master
or, even better, backups could be named also by the change-id
(if they are not yet) and a "logerrit restore <change-id>" could be
more user-friendly. Also, some logerrit support for identifying and
removing no more needed backups would be nice.
* Gerrit mails. I understand that there's some work ongoing on this, my
impression at that time was that they were not so clear, too much text
without useful information.
Also the flow of action was a little bit confusing:
1. I submitted a patch and a fix for it (patchset #2)
2. Then I received a notification about a new patchset applied by
"gerritbot", with no clear information on who changed what and
why, the only way to understand this was to have a look a the
diffs (on 72 files, realizing just on a second time that I needed
to manually select the base for the diff somewhere higher on the
webpage)
3. Then a notification that the patch was committed
4. Finally a comment that there were some minor optimizations that
could be applied. I agree that this could be useful, but at first
I found myself a little bit puzzled on which was the way to go. I
proceeded with a new standalone patch.
Apart from these quirks, that hopefully can be solved, in my opinion the
flow is quite usable and, maybe, a little more familiar than the
mail-based one.
HTH
Mattteo