On 29/07/2012 14:42, Petr Bena wrote: Need more Chad's :L When is this not being discussed/debated on/argued over... It took a while to checkout the SVN as well in my experience... especially if you checked out from trunk/phase3 or the whole of trunk... The guide used at the Berlin Hackathon this year was pretty good, and I managed to work it out, with the exception of jumping the gun a bit and using the wrong repo (key point: check what your doing) but other than that, I've managed to use Git everytime I've wanted to now. The guide for anyone that hasn't seen it, https://www.mediawiki.org/wiki/Git/Tutorial . As far as I remember, I've had no problems using that and setting up my git stuff a few times on Linux and once with TortoiseGit although I haven't tried pushing from that yet - although pull and clone seemed to work well. I'd say I kind of agree with the other points, but I'd also sort of disagree. Being someone that doesn't commit/push a lot of code, the things I have done (which, are relatively simple) we're easy for me to do... -- Lewis Cawte

On 29/07/2012 15:45, David Gerard wrote: I believe, although I may not be correct, Sumana requested these a while back and has been looking at them... I'm not sure if it covers (new) volunteer contributors or not. I think one thing that if it isn't a positive statistic is that from my point of view is that the volunteer contributors that we're out there in the SVN days and the ones that actually committed regularly have not yet learnt or are not as willing to relearn how to contribute... I know from some small experiences a few that are unhappy or unwilling that they have to relearn things. That, or Git, or the whole new workflow I assume, but who perhaps does this make it easier for? I suppose its easier especially for the people on the WMF side, since there is no more "submit patches and have them reviewed before you get commit" and its more get commit access, and then have your patches reviewed without them potentially breaking things. -- Lewis Cawte

On 29 July 2012 16:01, Lewis Cawte <lewiscawte(a)googlemail.com> wrote: I was just asking for numbers. Your response looks very like making an excuse for bad numbers *before we have the numbers*. I submit that if we know we have to prepare excuses for bad numbers before we even have the numbers, the current process may not have been a good idea. - d.

Since the operations repos have been opened, we've had over 800 changes pushed in from non-operations team members (this includes devs that have root). That's roughly equivalent to the number of changes one operations team staff members have pushed in during the same time period. So, ignoring numbers for all other repos (whether they are good or bad - I have no clue for those), the switch to git/gerrit has had a major impact for the operations team. - Ryan

Yes it _is_ difficult for volunteer developers. I still find it very difficult and did not commit any new line of code to gerrit except a "coached fix" during Berlin Hackathon 2012. And use github for my daily work now. Tom

Am 29.07.2012 21:56, schrieb Ryan Lane: The previous committers to this thread already said it all. Basically, I miss exspecially these functions which I often used in the CodeReview - list a code-(module) related contributions of committers - immediate and direct view of accumulated differences (this was so nice) - show differences in a module between this day and that day (for bisection of a certain bug) I mean this: just as an example https://svn.wikimedia.org/viewvc/mediawiki/trunk/extensions/OpenID/OpenID.p… Revision 115069 - (view) (download) (annotate) - [select for diffs] Modified Thu Apr 26 23:50:06 2012 UTC (3 months ago) by wikinaut File length: 10965 byte(s) Diff to previous 113562 , to selected 111329 - clean interface not to bother with links (in gerrit) I do not understand and I do not have to use. Summary: I as a now-and-then contributor need a "gerrit-light" with a interface very much the same as "CodeReview". This is why I proposed already in another mail to write a wrapper, a gateway or however you want to call it which wraps gerrit, and let me use gerrit as I was used to do with CodeReview. Expert will switch away fro "gerrit-light" to "gerrit-expertview" of course. Tom now also on #mediawiki

On Sun, Jul 29, 2012 at 10:39 AM, Lewis Cawte <lewiscawte(a)googlemail.com> wrote: No, we need a real process for this that scales. I was going to start some work on this with a new extension called GerritTools that had a "request queue" to automate much of this--but I got sidetracked from it. If someone's looking for a project... ;-) -Chad

2012/7/29 Petr Bena <benapetr(a)gmail.com>om>: I am really not a fan of Gerrit, but this is an exaggeration. What is true, however, is that the documentation is still not great. There is [[Git/Workflow]], which is 30 pages long. [[Git/Tutorial]] was supposed to make it simpler, but it grew to be 36 pages long. Based on the idea that describing hard things on one page with useful links can make these things easier, I just attempted to write this: https://www.mediawiki.org/wiki/Git/TLDR Hopefully some people will find it useful. Feel free to edit it, but don't let it grow longer than one (printed) page. -- Amir Elisha Aharoni · אָמִיר אֱלִישָׁע אַהֲרוֹנִי http://aharoni.wordpress.com ‪“We're living in pieces, I want to live in peace.” – T. Moore‬

On Mon, 20 Aug 2012 09:15:00 -0700, Sumana Harihareswara <sumanah(a)wikimedia.org> wrote: It also looks like it didn't even go on for long. I'd like to know what kind of issues he had, rather than just the fact that he had issues, and we know there are issues which may or may not be the ones he had trouble with. He's using GitHub so it looks like he's fine with the Git part. Is it Gerrit he had trouble with or the fundamental idea of Gated repos where he waits for review? Or is his issue more with the current trouble trying to create new repos and get into development? If he has Gated repo issues, would Gerrit's direct push permission work? If Gerrit is really the issue and he doesn't actually have an issue with the idea of a Gated repo, then would another gated workflow work? eg: Would Phabricator's work? Or would mine work? -- ~Daniel Friesen (Dantman, Nadir-Seen-Fire) [http://daniel.friesen.name]