I give up on getting CircleCi to behave, so this is just good for us devolpers to run the same tests

Codecov Report

Patch coverage: 15.24% and project coverage change: -0.54 ⚠️

Comparison is base (c3d2e62) 39.73% compared to head (7220bde) 39.20%.

❗ Current head 7220bde differs from pull request most recent head 1c5d0b4. Consider uploading reports for the commit 1c5d0b4 to get more accurate results

@@            Coverage Diff             @@
##           master    #8682      +/-   ##
==========================================
- Coverage   39.73%   39.20%   -0.54%     
==========================================
  Files         290      294       +4     
  Lines       50374    51365     +991     
==========================================
+ Hits        20017    20138     +121     
- Misses      30357    31227     +870     

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

Btw, the segfault bug appears to be only with 18.04, so we can leave the appimage on 16.04 for max compat and update the rest of the docker images to 20.04 which is nice and 2 years old by now :) This way you can still keep docker

Really? I'd swear 20.04 is one of the very first things I tried and that I ran into the same issue. That'd be good news. :-)

It was working on GHA and 18.04 wasn't, so i assumed it might be the same for docker

If it does work and we upgrade I would like to help clean up the docker image: Install everything in system directories and get rid of the need to mangle paths

You can do that regardless if you want. Docker will remain in use. What do you mean mangle paths though?

every Ci script has to add the user luarocks directory to its path and the shellcheck path etc

If it "has to" then that's an oversight due to that being required before Docker.

If it "has to" then that's an oversight due to that being required before Docker.

most of the cruft is not from docker rather from how the CI scripts were originally set up from travis and before

@yparitcher Maybe it would make sense to add a make doc check too. See #8742 and #8636 (comment).

Hopefully this could help to shortcut some PR's.

Btw, I don't currently run ldoc on PRs because it almost always wastes about 10 additional seconds. A mistake that slips through just means the docs will be a couple of days out of date at worst.

@yparitcher Please also add

for skipping checks onf shfm

Thanks.

Btw, I don't currently run ldoc on PRs because it almost always wastes about 10 additional seconds. A mistake that slips through just means the docs will be a couple of days out of date at worst.

I also haven't done this regularly.. So this is no big deal. B u t; If we are at the hunt, if we are on a special thing, if we are onto it: It might be easy to solve right now. And 10s now is better than 100s the next day :) :) :) ;)

It's a very deliberate choice. People who add docs normally test them first to see if they look okay! Otherwise that's like submitting a PR with untested code. There's an expectation that it works at least in the emulator. ;-)

Anyway, point being it's a giant waste of time and the risk of failure is zero.

Is this PR still relevant?

Ah, forget what I just said. Let's close them both and we'll find it easily enough in the future. :-) I didn't even realize/remember there were two PRs.