Package Details: chatterino2 2.5.4-5

Git Clone URL: https://aur.archlinux.org/chatterino2.git (read-only, click to copy)
Package Base: chatterino2
Description: Second installment of the Twitch chat client series "Chatterino"
Upstream URL: https://chatterino.com
Licenses: MIT
Conflicts: chatterino
Provides: chatterino
Submitter: DodoGTA
Maintainer: DodoGTA
Last Packager: DodoGTA
Votes: 2
Popularity: 0.017974
First Submitted: 2024-06-06 17:20 (UTC)
Last Updated: 2025-11-30 21:50 (UTC)

Required by (3)

Sources (17)

Latest Comments

1 2 Next › Last »

Eidokan commented on 2025-12-03 12:24 (UTC)

The git config in use inside of the container is installed in mkchrootpkg line 193, function called in line 388 / devtools version 1:1.4.0-3 and contains just

[safe]
    directory = *

This is likely to avoid issues related to the user used for building ("builduser" inside the container) not owning the source files.
I don't see how no config vs those specific keys "missing" could be relevant. What I can tell you is, that there is a repository-local, a user-wide and a system-wide config plus some sort of auto-detection that I have only ever seen failing. You could try to run something like git show --pretty=format:"Author: %an <%ae>%nCommitter: %cn <%ce>" HEAD | head -n2 on the repository as left behind by the build process to see author and committer of the commit. Maybe that will shine some light on where the user information comes from in your setup.

DodoGTA commented on 2025-12-03 09:21 (UTC)

@Eidokan I tested my commit fix with a missing user.name and user.email and it worked fine (so maybe the issue only occurs on a completely empty or missing gitconfig?)

Eidokan commented on 2025-12-03 00:40 (UTC)

The changes worked perfectly fine for me. I really didn't expect you to patch the ports, I just wanted to share what my issue was, since it was not immediately obvious for me. So thanks a lot for that.
If we are still nitpicking about things that could be improved, the git issue I mentioned is the only thing that comes to my mind. git commit --no-edit fails in case people didn't configure name and e-mail, which is always the case inside of build containers as used by devtools or simply if git got installed automatically or is for some other reason not configured. It can of course be set just for that repo or overwritten inside of prepare(), however I still don't know about any convention about what to put as an e-mail when there really isn't one. Anyway if someone wonders how to do that for just that one command, I used

git -c 'user.name=chatterino2 PKGBUILD' -c 'user.email=invalid@chatterino2.aur.archlinux.org' commit --no-edit

I am, again, not sure if that one justifies an update. People affected by that are probably capable of working around that themselves.
Thanks a lot for your efforts!

DodoGTA commented on 2025-11-30 21:53 (UTC)

@Eidokan I pushed another (and hopefully final) update to resolve the remaining build issues (but with my relatively complex build sequence, something always goes wrong in some scenarios)

Eidokan commented on 2025-11-28 12:22 (UTC)

I figured it out. The ports used by the test were partially in use by tor. 9050 is the default port for their socks proxy and 9051 their default control port. In that log, the former port was listening, the latter was not. I should have thought about that earlier. When I stop that daemon, the tests pass. Unfortunately those ports appear to be hardcoded in chatterinos test suite. Welp, now I know…

Eidokan commented on 2025-11-28 12:06 (UTC)

Well thats embarrassing. I uploaded new logs and linked the old ones from -1. Here is the new snippet. Sorry for that. And I agree, it's weird. I will do some testing.

DodoGTA commented on 2025-11-28 08:03 (UTC)

@Eidokan It's strange that test still fails (considering it now passes on my system)

Maybe you should upload a log of the failed test?

Eidokan commented on 2025-11-27 20:12 (UTC)

Hi, I just build the new version (2.5.4-4) and the Qt thing worked fine. The tests however did not and - without the "or true" - caused check() to fail. It is down one failure though ("test 70 - PluginTest.testHttp (Failed)" works now).
Unfortunately the git commit command in prepare() triggered another nuisance with mkchrootpkg from devtools. That script unconditionally installs a git config without user and email entries inside of the container causing git commit to print its "tell me who you are" message and fail. I'm not sure if there is some AUR/makepkg convention on how to deal with git not being configured, so I just overwrote those settings with some nonsense to work around that.
I also added the ctest exemption back to get a full log and a successful build. Most importantly though, the resulting package still works!

DodoGTA commented on 2025-11-27 12:34 (UTC)

@Eidokan I just pushed the headless workaround (and some other improvements)

Hopefully they don't cause issues

Eidokan commented on 2025-11-05 15:19 (UTC)

Oh and thank you, pajlada, for your work on chatterino itself! I completely missed you being here

1 2 Next › Last »