So I looked in the atom ebuild, and it's huge. It also says it was generated by a python script in the atom overlay. So I looked and that script and it's huge, and also atom-specific. So I think an eclass should be made to facilitate packaging general electron apps "correctly".
One reason the script is there is to find out all the binary dependencies by going through the dependency tree. It's a ton of tedious work to do manually and a eclass wouldn't really help with that. Another reason why I think going eclass is a bit premature is that app-editors/atom is so far the only package that depends on electron, so it's hard to identify and extrapolate the commonalities with other such packages.
(In reply to Elvis Pranskevichus from comment #1) >a eclass wouldn't really help with that. Really? > Another reason why I think going eclass is a bit premature is that > app-editors/atom is so far the only package that depends on electron, so > it's hard to identify and extrapolate the commonalities with other such > packages. The only electron app in existence or the only one in the tree? I just packaged another electron app, but all it does is dump the whole thing in /opt. And I see a third one I might package someday.
If someone(In reply to Nathan Dehnel from comment #2) > The only electron app in existence or the only one in the tree? I just > packaged another electron app, but all it does is dump the whole thing in > /opt. And I see a third one I might package someday. In such case you are welcome to help out with writing the electron.eclass.
I second Nathan's motion for an "electron.eclass". As Atom ebuilds demonstrate, packaging Electron-based apps is absurdly non-trivial. As Electron goes mainstream, an official eclass would reduce everyone's maintenance burden. In my case, I'd like to package HakuNeko (https://github.com/manga-download/hakuneko), an Electron-based manga aggregator. It's unclear where exactly to start, though. I'll probably end up defaulting to Nathan's ad-hoc "solution" of stuffing everything into /opt -- which, of course, runs contrary to the Gentoo ethos in every possible way. An Electron eclass might have been premature in mid-2018, when this feature request was first filed. Today, in late 2019, a proper eclass is all but mandatory for sane packaging of this popular ecosystem. Electron isn't going anywhere. Let's get this right as early as possible and save everyone a heavy load of bitrot.
