Will Thompson

I am about to find a similar bug about the behaviour of eol-rebase'd apps with `--no-deploy`, which leads me to think that making uninstall ops do nothing if `no_deploy` is...

Great point. I hadn't considered the case of sandboxed apps accessing thumbnailers from other sandboxed apps. I think it would be sad to declare that the file manager can only...

I haven't thought this through for more than about 2 minutes, but you could imagine Flatpak exposing the installed thumbnailers (host or Flatpak:ed) to sandboxed apps in a controlled way.

I tried to use `aioresponses` https://github.com/pnuckowski/aioresponses but I am apparently not smart enough to make this work. The mock object gets stopped too early for reasons I don't understand.

You can use the json checker as a workaround. Try: ```json "x-checker-data": { "type": "json", "url": "https://api.github.com/repos/ArtifexSoftware/ghostpdl-downloads/releases/latest", "version-query": ".name | split(\" \") | .[1]", "url-query": ".assets[] | select(.name|test(\"^ghostscript-(.*).tar.gz$\")) | .browser_download_url"...

The GNOME checker supports GNOME's (previous) similar scheme. It does seem a reasonable addition elsewhere. In the linked case of Calls you could fake it with: ```yaml x-checker-data: type: git...

It's true that in other cases the tarball name gives a better message but I do agree that there is scope for improving the commit message(s) and/or PR message. Patches...

I've hit the same with the CLI Vim from `org.vim.Vim`.

I wonder if there's something clever that could be done like, “is attached to a PTY which is in a GNOME Terminal window”?

I agree that this would be interesting to try! One note: right now the way the bot avoids opening the same pull request more than once is that it derives...