Invent/gitlab, issues and bugzilla

[Boudewijn Rempt]

On donderdag 4 juli 2019 20:42:31 CEST Martin Flöser wrote:
> 
> Boud pretty much describes the problem large projects (krita, kwin, 
> plasma) have with bugzilla. We don't use bugzilla to handle bug reports, 
> but to somehow manage all the reports we get, to survive. I blogged 
> about these issues years ago, I did statistics for KWin showing that > 
> 90 % of the incoming bug reports are not bug reports, but rather 
> duplicates, user support, etc. The situation hasn't changed much over 
> the years. Reading Boud's comments it looks like the situation became 
> much worse for krita lately and I do not want to swap with him.

For the past six months or so, Krita is the KDE product which gets the most bug reports. We're also seeing a continuation of the real rise in downloads, social media traffic and user base. That results in lots more interaction with end-users, which I normally like. I like my users. The big difference between working on Krita and all other jobs I've ever had is that when I work on Krita, my work ends up being used.

> Obviously for any large project the idea of swapping to an inferior 
> system (which gitlab seems to be, haven't used it yet) is a horrible 
> idea. But that's not the problem. The problem is that users interact 
> with developers directly. The user goes to bugzilla or gitlab issues and 
> reports a support request. Instead of friendly user support he gets a 
> grumpy third level support answer that this is not an issue. No help, 
> frustrated user and frustrated dev who just wasted another minute on 
> bugzilla to handle user support.

I try not to be grumpy... But we've closed ask.krita.org and I've retired from reddit and we don't do support on Twitter -- because it's no longer manageable.

> No company would allow to have Boud handle user support. That's insane. 
> He's product owner, chief technologies, call him whatever you want, but 
> he is not first level user supporter.

In fact, I hired one of the volunteers who spent all her time helping people to help me fix bugs. It's working out great, and I've hired a second bug-fixing minion, and now the first minion (they choose their titles) is trying to help to figure out how to plug the user support gap.

> That's the problem we have to face. Whether we use bugzilla or gitlab 
> issues for internal issues is irrelevant. 

Not totally irrelevant. Gitlab just does not cut the mustard for tracking bugs. It was not designed for that, it cannot handle that. The whole discussion about ditching bugzilla and moving to gitlab issues is, as far as I am concerned, a has-been. It's over. You don't point people at something like invent.kde.org and ask them to figure out how to report a bug.

> What is important is that we a
> get users to use an adequate user support forum and not bugzilla or 
> gitlab issues. Once we get the shit reports out of our bug tracker, 
> everything looks different.

I'm not very hopeful about that happening, though.

> Right now from KWin perspective I would say "hell no!" to the idea of 
> using gitlab issues. But a better tool for development which gitlab 
> issue seems to be is something I would like to have. Because for 
> development planing I never really liked bugzilla. Although it might be 
> that it was just too useless due to all the invalid bug reports.

Yes, development planning is not what bugzilla is for. We need a four-stage system:

1. User support system. User reports an issue, gets a canned answer in 90% of the cases. In 10% of the cases, they have found a bug.
2. Bug tracker. Those user bugs, and the bugs the developers find are registered. With the right component, version, all the data we can get. Once the bug gets assigned, we move to system 3.
3. Task system. Information about one or more bugs gets consolidated into one task, and a plan is formulated. The task gets updated with commits, comments and so on during the implementation phase. It ends in  review request or merge request...
4. There's a merge request or review request, and once it gets accepted or merged, 3 is informed, 2 is informed and 1 is informed.

> My suggestion is to address the user support and then look into whether 
> we want to keep bugzilla or switch to gitlab issues.

For me it's clear: gitlab issues can, barely, replace Phabricator's tasks. They cannot replace a bug tracker. 

-- 
https://www.valdyas.org | https://www.krita.org