USE codes for firefox-bin[SOLVED]

figueroa · Last edited by figueroa on Sun Nov 22, 2020 5:26 am; edited 1 time in total

I find that firefox-bin works better across sites than the firefox I build myself. Is there a way to know what USE codes are being used to build firefox-bin?

Alternatively, what USE codes (other than default for 17.1 desktop profile) are users finding the most happiness with?
_________________
Andy Figueroa
hp pavilion hpe h8-1260t/2AB5; spinning rust x3
i7-2600 @ 3.40GHz; 16 gb; Radeon HD 7570
amd64/23.0/split-usr/desktop (stable), OpenRC, -systemd -pulseaudio -uefi

fedeliallalinea · Posted: Fri Nov 20, 2020 7:07 pm Post subject:

firefox-bin is official binary release, it's not like libreoffice-bin because you can't build a firefox version and redistibute it with firefox trademark.
_________________
Questions are guaranteed in life; Answers aren't.

Ionen · Developer Joined: 06 Dec 2018 Posts: 2716

Do you mean what you'd need to somewhat replicate firefox-bin with your own build?

As pointed above it's upstream's release as-is, but it notably use clang + lto + pgo (or so I've read anyway).

GCC builds being hardly tested by mozilla often brings regressions, "if" you are getting obscure segfaults it's often related. Gentoo does go out of its way to fix build failures but fixing runtime issues is hard, I strongly suggest USE=clang myself.

Performance-wise pgo should be much more significant than what -march=native would give you, so if you build with USE=-pgo you already have a worse firefox than firefox-bin (but pgo doubles build time.. so if that's a problem you may as well use firefox-bin in my opinion, unless you need other build-time changes -- personally I do custom patch firefox so..

).

If want to use firefox's vaapi, that "currently" requires firefox to be built with USE=wayland (even on x11, or else vaapi code doesn't get built in). Not an issue with firefox-bin because it's built with wayland support in the first place even if you USE=-wayland that won't break it.

If you need alsa, it still works but firefox's native alsa is suffering from bit rot and unusable on a few sites, so like firefox-bin you're often better off using apulse (either through apulse[sdk] or the ebuild's workaround).

firefox-bin does use less system libraries but there's little sense in copying that behavior.

On a side-note, like -march, you can use RUSTFLAGS="-Ctarget-cpu=native" which firefox-bin wouldn't use (there's no need to specify the opt-level in RUSTFLAGS, appropriate one will be added).

figueroa · Posted: Fri Nov 20, 2020 7:50 pm Post subject:

Ionen · Developer Joined: 06 Dec 2018 Posts: 2716

When it comes to odd runtime issues I tend to suspect the lack of USE=clang, also not entirely impossible it's related to system-* libraries but I doubt it.

I'd like to tell you if I'm affected by the same issue (with my firefox 83 built with clang), but I have no idea how to reproduce especially if it's something that takes a while to happen. I haven't had any firefox issues in over 2 years though.

I doubt most of my flags have relevance on this but using:

figueroa · Posted: Fri Nov 20, 2020 9:04 pm Post subject:

Etal · Veteran Joined: 15 Jul 2005 Posts: 1931

You can see and compare the build options if you go to about:buildconfig in Firefox.

Of course you'd need to map them back to USE flags.

figueroa · Posted: Sat Nov 21, 2020 6:19 pm Post subject:

I've rebuilt firefox 78.5.0 several times, now with USE +clang -dbus -openh264 giving me:

Ionen · Developer Joined: 06 Dec 2018 Posts: 2716

Personally not getting any console warnings playing youtube with built firefox-83.0 (or at least not with the videos I tried), console is entirely silent (checked in-browser console too, but nothing similar).

Have you been comparing with firefox-bin:esr/78 now or only 83.0? Just to be sure if it's not ESR78 vs 83.0

That aside, note that USE=pulseaudio is needed to use apulse with built firefox, it enables the native alsa bit rotting support if you disable it. Will need to install pulseaudio anyway because need pulseaudio headers to build firefox but then use apulse at runtime if detected.

figueroa · Posted: Sat Nov 21, 2020 6:48 pm Post subject:

Ionen · Developer Joined: 06 Dec 2018 Posts: 2716

Yeah currently that needs rust-1.47.0 (which can use llvm11, 1.46.0 is pinned at 10), but I kinda doubt lto/pgo have anything to do with it.

If it's media playback-related I guess it's "possible" apulse vs native alsa is related though. From previous threads "think" I remember you have a rather intensive alsa setup and firefox's half broken native alsa support may not necessarily play well with it.

figueroa · Posted: Sat Nov 21, 2020 7:24 pm Post subject:

figueroa · Posted: Sat Nov 21, 2020 8:32 pm Post subject:

Finding: I get a similar error abruptly stopping a Youtube video running firefox-bin from the commandline. Only thing is that I don't experience the usability issues with firefox-bin. Unlike source built firefox, firefox-bin has been steady as a rock.

This is the error tossed while running firefox-bin:

figueroa · Posted: Sun Nov 22, 2020 5:25 am Post subject:

Late this afternoon, I upgraded firefox to www-client:firefox-83.0 ~amd64. I've been running it for hours with no apparent usability problems. After testing with numerous Youtube videos, Facebook and publix.com, the two sites that would not resolve correctly after watching Youtube videos, continue to work predictably.

I'll mark this solved, though I don't like running major software as ~amd64.
_________________
Andy Figueroa
hp pavilion hpe h8-1260t/2AB5; spinning rust x3
i7-2600 @ 3.40GHz; 16 gb; Radeon HD 7570
amd64/23.0/split-usr/desktop (stable), OpenRC, -systemd -pulseaudio -uefi