hwaccel on firefox still relevant?

[logrusx]

Today I finally ditched firefox:rapid in favor of the version from the binhost. However it prompted me I had to disable hwaccel use flag to make it compatible with the binhost. I opened the ebuild and saw that the only thing that hwaccel pulls in are two packages libva-utils and pciutils none of which seems to be a library, which I conclude means firefox does not link against anything in addition to what it already links to. The rest related to hwaccel are settings, which I think one can set manually if they want. Also the referenced Mozilla bug 594876 is closed in favor of webrenderer.

So, does hwaccel make any difference aside from the settings it imports into the browser? If not, does it make sense to file a bug prompting developers dropping that flag might actually be a good idea?

EDIT: Maybe I should clarify that I'm considering this from the perspective of binary packages. Hwaccel adds those as runtime dependencies, yet a binary package will refuse to install if it is enabled. Also it only imports settings so I guess it doesn't matter if this use flag matches the binary package's one.

Best Regards,
Georgi

[alamahant]

Yes but libva-utils pulls libva if it is not already pulled by USE="vaapi".
And hardwareaccel seems to work fine in firefox.
_________________

[logrusx]

[Banana]

well you can try it with a youtube 4k-8k video and compare the CPU usage while playing.

And If I read https://wiki.gentoo.org/wiki/Firefox#Hardware_acceleration correctly the useflag trigges the need of vaapi but can be disabled. If those helpers are installed it should work, too.
hmm...
_________________
Forum Guidelines

PFL - Portage file list - find which package a file or command belongs to.
My delta-labs.org snippets do expire

[logrusx]

Whether it works or not is not relevant to the question.

[Juippisi]

Yes, the use flag will install vaapitest/v4l2test binaries which firefox uses to self-test hardware acceleration and provide debug help in case it doesn't work. These were originally installed unconditionally, but some people without (working) hardware acceleration had problems launching firefox due to that. Although I don't remember the specifics, it could be they had something else weird going on, but the problem was immediately fixed by rm'ing the vaapitest binary.

1: IIRC the binhost builds multiple variants of firefox, so you could request +hwaccel to be one of them. Luckily firefox isn't that heavy, even with +pgo so I believe the request is fair. Although don't do this yet, read further...
2: I'd be curious to know if hwaccel works without the binary? Are you certain it's working simply by changing settings and emerging the missing deps?
3: maybe you can try to grab the vaapitest binary from firefox-bin and drop it as /usr/lib64/firefox/vaapitest :)
4: 134.0 is out next week, maybe vaapitest binary can be installed unconditionally again and we investigate a bit better with people having problems with that. Since firefox-bin shouldn't then work either, although that's also not unheard of when/if using older hardware.

hwaccel in firefox is a bit unfair as a use flag, especially if the binary can be solved. As suggested by sam, we should implement a package called "firefox-gentoo-configuration" that does all about:config changes so none of those need to be in the main firefox ebuild. Then that config package can have 'hwaccel' use flag.

[logrusx]

[logrusx]