Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#77225 - Set CAP_SYS_NICE on sway binary
Attached to Project:
Community Packages
Opened by Dom DiMaggio (Damino) - Sunday, 22 January 2023, 01:45 GMT
Last edited by Brett Cornwall (ainola) - Saturday, 04 February 2023, 00:00 GMT
Opened by Dom DiMaggio (Damino) - Sunday, 22 January 2023, 01:45 GMT
Last edited by Brett Cornwall (ainola) - Saturday, 04 February 2023, 00:00 GMT
|
DetailsDescription: /usr/bin/sway needs cap_sys_nice=eip set to avoid avoid stuttering in certain scenarios.
Additional info: * package version(s) 1:1.8-3 * link to upstream bug report, if any Support added here: https://github.com/swaywm/sway/pull/6994 Expectation from developer for package maintainers to set stated here: https://github.com/swaywm/sway/pull/6992#pullrequestreview-972760643 Steps to reproduce: GPU-limited gaming (Final Fantasy XIV via Wine) on one monitor while having moving content open on a secondary monitor (e.g. video playing on Firefox). My game frequently stuttered until I ran `# setcap 'cap_sys_nice=eip' /usr/bin/sway`. The same could not be observed with the same scenario using GNOME or Plasma Wayland sessions. A quick look into both determined they request SCHED_RR differently from Sway and therefore don't require this. |
This task depends upon
Closed by Brett Cornwall (ainola)
Saturday, 04 February 2023, 00:00 GMT
Reason for closing: Implemented
Additional comments about closing: sway-1:1.8-5
Saturday, 04 February 2023, 00:00 GMT
Reason for closing: Implemented
Additional comments about closing: sway-1:1.8-5
Comment by Toolybird (Toolybird) -
Sunday, 22 January 2023, 21:53 GMT
If this is something packagers should be doing, they should document it properly, instead of it being buried away in some comment or changelog...
Comment by Brett Cornwall (ainola) -
Sunday, 29 January 2023, 07:12 GMT
@maximbaz, this seems reasonable to me. Any thoughts?
Comment by Maxim Baz (maximbaz) -
Monday, 30 January 2023, 20:23 GMT
Sounds good to me, please go ahead if you have time to do it :)
Comment by Brett Cornwall (ainola) -
Wednesday, 01 February 2023, 05:48 GMT
Damino, I've pushed sway-1:1.8-4 to [community-testing]. I'll do some testing when I have some time but I'd love your feedback as well!
Comment by q rty (q234rty) -
Wednesday, 01 February 2023, 09:39 GMT
This needs https://github.com/swaywm/sway/commit/2f2cdd60def006f6d3cbe318f9edd7d68fcb239a otherwise this will break user xkb configs.
Comment by Dom DiMaggio (Damino) -
Wednesday, 01 February 2023, 12:31 GMT
sway-1:1.8-4 is working for my use case now, thank you! Unsure about q234rty's comment, as it's a feature I don't use.
Comment by Brett Cornwall (ainola) -
Thursday, 02 February 2023, 06:25 GMT
Thanks for letting me know about the xkb configs, q234rty! I've pushed sway-1:1.8-5 to [community-testing]. Can you test/verify that this works for you?
Comment by Dom DiMaggio (Damino) -
Thursday, 02 February 2023, 12:27 GMT
Just writing to confirm my issue is still resolved with sway-1:1.8-5.
Comment by i0f (I0F) -
Friday, 03 February 2023, 09:22 GMT
Right now the community-testing PKGBUILD uses 'setcap cap_sys_nice+ep usr/bin/sway', while in the inital bug report 'eip' is used instead of 'ep'. Is one better than the other?
Comment by q rty (q234rty) -
Friday, 03 February 2023, 12:43 GMT
I can confirm that sway-1:1.8-5 works with user xkb configs while sway-1:1.8-4 doesn't.
Comment by Brett Cornwall (ainola) -
Friday, 03 February 2023, 17:14 GMT
i0f: Good question! The inheritable set needn't be specified as sway is the binary requiring the capability, not any other spawned processes/children.