Hello list!
Gaba and I recently talked about how it would be good to have some status about where we are with the Gitlab migration project.
In December a lot of you reported a lot more issues than I had spotted, which was both extremely good, but also not so good in that every issue was another thing for me to fix :-) Thanks to everybody for reporting these!
Fortunately, with anarcat's suggestion of using some parts of tracboat I got over some of the "Trac markup" to Markdown issues quickly, but some of them remained.
Recently, cohosh reported an issue with the Gitlab instance we were running where PR's generated 503 messages. Hiro spend a lot of time trying to debug this, and with this new issue in mind, we have decided to move to Gitlab's own "omnibus" package method for maintaining the Gitlab instance instead of Debian's "salsa" Ansible playbooks. We feel that a larger number of us are more confident with operating the omnibus Gitlab instance and those of us who have been running Gitlab instances this way before (hiro, gaba, and myself) haven't run into issues like the ones we have had with Salsa so far.
Here's a summary for my Gitlab hackday today:
- Synced my local cache of Trac tickets (used for speeding up runs). - Worked on "missing information" issues reported in December: points -> estimation, actual points -> spending (both based on a conversion function Gaba and I talked about in December), owner, milestone, reviewer, and "reported by". I still use my own username for these to avoid spamming folks with emails. - CC in issues for *known users* so far. That is the people who have users on GL when we move. - Figure out how to handle parents. - Extended the "Tracboat" conversion code with handling some of our {{{ }}} (code blocks) and headlines better. Checking the cosmetic changes for errors takes longer time than all the metadata issues since I need to go over tickets and see if things "looks right" and not just look out for missing or wrong information.
Right now it's running against my local GL install in a Qubes VM for speed, and once I've gone over the results I'll make it post to a new repo on dip.torproject.org (this makes the difference of a sync from 8h for a complete sync to 20min). I will be using "strike through" on the text on our December pad to indicate that I think something is fixed, then we can walk over it together at a later point.
Have a nice weekend all!
All the best, Alex.