I really wish I could bring myself to get into sxmo more. It seems like something I'd like, but I mostly find myself confused when I try to use it. Even if it's not perfectly suckless, maybe it's a bit more in the "suckless" dimension than I can bear--I did like dwm for a while, but I switched to i3wm and later Sway and never really looked back.
There's obviously a lot of small little projects that implement cool ideas, but one thing that is a little bit of a bummer is that there's really no obvious solution that you can flash onto a Pinephone and call it daily-driver ready. It would really be nice if standard-ish Linux desktop environments could be adapted to work well on phones; I mean, at this point, the proof-of-concepts are enticing enough for me to believe that it'd be worth the effort. That having been said, I've been wondering if maybe to get the Pinephone to a working state, if it'd be better to actually go for a very minimal base system and try to build a more or less non-standard usermode. Something like, musl, pipewire, eg25-manager, a custom wlroots compositor that implements most of the actual phone features directly, and some simple system software to go under it (file browser, terminal, etc.) The main thing I really want is to get the absolute best possible battery efficiency, something that can manage rtcwake to occasionally check for notifications and handle some basic alarm clock functionality, and intelligent enough system software to e.g. restart the EG25 when it seems to be stuck. (Usually on Phosh + Debian, restarting the eg25-manager systemd unit is enough, so apparently it'd be good enough to just find a way to detect when it's stuck and restart eg25-manager.) It's a lot of work, and I admit that it feels like you'd be going a tad in the direction of Android by ditching most of the standard userland. But on the other hand, there's so many damn projects involved with most functionality in the device that it is a bit difficult to even know where to start when it comes to troubleshooting, and in my opinion it'd still be nicer than Android if the userland was "standard" enough to still run typical desktop apps and run more-or-less stock kernels.
Then again, for now, I feel my frustration would be better spent trying to debug what's already there. I'm wondering if maybe it would be possible to improve the wake times, for example... I have a sneaking suspicion that most of the resume time is taken before the Linux kernel gains control, but maybe it'd be worth trying to get something like pmgraph running to see if there's any room for improvement.
There's obviously a lot of small little projects that implement cool ideas, but one thing that is a little bit of a bummer is that there's really no obvious solution that you can flash onto a Pinephone and call it daily-driver ready. It would really be nice if standard-ish Linux desktop environments could be adapted to work well on phones; I mean, at this point, the proof-of-concepts are enticing enough for me to believe that it'd be worth the effort. That having been said, I've been wondering if maybe to get the Pinephone to a working state, if it'd be better to actually go for a very minimal base system and try to build a more or less non-standard usermode. Something like, musl, pipewire, eg25-manager, a custom wlroots compositor that implements most of the actual phone features directly, and some simple system software to go under it (file browser, terminal, etc.) The main thing I really want is to get the absolute best possible battery efficiency, something that can manage rtcwake to occasionally check for notifications and handle some basic alarm clock functionality, and intelligent enough system software to e.g. restart the EG25 when it seems to be stuck. (Usually on Phosh + Debian, restarting the eg25-manager systemd unit is enough, so apparently it'd be good enough to just find a way to detect when it's stuck and restart eg25-manager.) It's a lot of work, and I admit that it feels like you'd be going a tad in the direction of Android by ditching most of the standard userland. But on the other hand, there's so many damn projects involved with most functionality in the device that it is a bit difficult to even know where to start when it comes to troubleshooting, and in my opinion it'd still be nicer than Android if the userland was "standard" enough to still run typical desktop apps and run more-or-less stock kernels.
Then again, for now, I feel my frustration would be better spent trying to debug what's already there. I'm wondering if maybe it would be possible to improve the wake times, for example... I have a sneaking suspicion that most of the resume time is taken before the Linux kernel gains control, but maybe it'd be worth trying to get something like pmgraph running to see if there's any room for improvement.
reply