Closed notpeelz closed 1 week ago
I'm thinking maybe it would be a good idea to rename SWAYSOCK
to SWAYFXSOCK
after all:
sway --get-socketpath
, it returns SwayFX's socket path, which could be considered incorrect, as sway isn't running.XDG_CURRENT_DESKTOP
is the preferred way to check what compositor the user is running; however, sway doesn't set that environment variable, which becomes the user's responsibility to set it to the correct value (and therefore might be unreliable). Renaming it to SWAYFXSOCK
gives programs a more reliable way to determine the "flavor" of sway in use.50-systemd-user.conf
, which updates SWAYSOCK
in the systemd environment block, breaking all scripts that make use of swaymsg
.SWAYSOCK
is pretty much the only thing remaining that could result in accidental "crosstalk" between compositors.Marking as draft until the review is done.
Edit: nevermind, draft PRs can't be reviewed. I thought it just blocked merging lol
pokes @WillPower3309
Could we provide distro packages like sway-is-swayfx
or swayfx-sway
to provide /usr/bin/sway{,msg,nag,bar}
? This could break some packages depending on the binaries or the sway
package name.
I've squashed the fixup commits and added a new subpackage to alias swayfx's versions of sway{,bar,msg,nag}
.
I also tried modifying the Fedora spec file, but rpkg was giving me strange errors (tested in a Fedora 39 VM).
Also, rpkg-util is no longer maintained/developed, so we should probably use some other packaging tool (such as fedpkg, dist-git or tito).
We usually use fedpkg
for day-to-day packaging works. https://docs.fedoraproject.org/en-US/package-maintainers/Package_Maintenance_Guide/#common_fedpkg_commands Hope this helps!
Thanks! I'll look into it.
BTW, if we go forward with the SWAYSOCK
rename, the compat packages could ship a config file to alias the environment variable, e.g.
/etc/swayfx/config.d/50-swaysock-compat.conf
:
exec systemctl --user set-environment \
SWAYSOCK="$SWAYFXSOCK" \
I3SOCK="$SWAYFXSOCK"
I rebased against the latest changes and got rid of most packaging-related commits. This should make it easier to review. I plan on submitting the rest of the changes in separate PR(s).
Hey peelz! Thanks a ton for all the work here and apologies on my end for the delays. I think it's a good call to save the swaysock change for a separate PR as I'd like to have more of a conversation there on if this is something we'd want
Just a minor comment and a question
Allowing side-by-side installing (where no binaries are named the same) with sway would also really ease packaging for debian and in consequence, ubuntu
Notes:
SWAYSOCK
wasn't renamed because it would break e.g python scripts using i3ipcswaynag_command
are unchangedsway(1)
) was updated, hopefully you will find it adequate :)CONTRIBUTING.md
was updated to reflect the direction of this projectPackages:
AUR: I've tested the AUR packages and use the PKGBUILD on my own machine as a daily driver. I have sway installed side-by-side with swayfx with no conflicts.
RPM: I'm not familiar with RPM. I've updated the spec file to the best of my ability. Untested.
Nix: Unchanged. I don't have a Nix/NixOS environment to test with.
Fixes #191