Speaking strictly as an end user who doesn’t have the skill or time to help in the necessary maintainer work, if I can’t play my Steam games and use Steam’s full capabilities on Fedora, then I will have to find a different distro where I can continue to do that. I don’t know how much, if any, weight that has on the scales here, but I felt I should chime in anyway.
I’m not sure its possible to fully capture what is worth someone else’s time in a project wide policy statement.
I would imagine for contributors who want to keep steam working, they probably have the right incentives to do the work necessary to keep a subset of 32bit deps working, so it becomes a question of the best place/way to do that work when others in the project want to stop doing the work of maintaining 32bit builds and want to lay it down.
So with that I have questions, in context of the proposal.
Key question:
how can we avoid blocking on that self-interest from a subset of the packagers while no longer asking the full group of packagers to carry the water of 32bit builds across the full project by default.
My first implementation question:
Is there any way for rpmfusion to provide the 686 builds for a set of steam deps, or does this need to be done in Fedora itself? Its unclear if this is even possible with how multiarch rpm works… im still in the process of rebooting my deeper rpm understanding from my previous life. But because we’re talking about steam, enablement in rpmfusion may be the best place. I just don’t know for sure if that’s possible. And Yes Neal, obs is a thing as well, I’m trying to talk to the most significant case presented, we can come back to other cases to see if any of these mitigations make sense as a general mitigation strategy.
My 2nd implementation question:
Does a steam enablement copr solve the “steam” problem adequately for those interested in keeping just enough 32bit deps built to keep stream running on Fedora? For something like derived linux distributions, this would not be an obvious change to their end users.
My 3rd implementation question:
If Neal is correct cocerning the Fedora infra that makes ELN branch possible, is there a way to proof of concept what that would look like for building just the deps 32bit steam (or obs needs?
Leigh Scott has mentioned that it is likely not possible to do all of this build work in RPM fusion.
Also less games work in the Flatpak and is not an official package from Valve, either, and thus also not officially supported. The native steam package is the only supported official means to run Steam.
I will put this in a language that probably will be more readable to anyone:
This change conflicts with the Fedora 2028 Strategy[1] by excluding the importance of gaming in the distro
The strategy state that we should grow the user base ↩︎
I think the idea is interesting. I just don’t think fedora is big enough to move valve. Canonical could not do it. And I don’t see Fedora as any different.
The main issue with the steam flatpak package and new users is two things basically.
A sandbox something that they’re not used to in the platform they come from and setting permissions makes no sense to them nor will they ever look up how to or try to understand it. They’re not here for that.
- Many gamers still have their libraries on NTFS based drives. They’re not going to change it to EXT4 they’re not going to re-download their libraries. It may be way more expensive to do that than anybody gives them credit for due to data and location limits.
What I mean here is you cannot read and write data without jumping through mini hoops of disabling the sandbox and even then you’re better off just using the native package for whatever distribution you’re running.
If you are a normal Linux user with some understanding of the platform. You can definitely work around these problems but if you’re not, you’re not going to.
The thing to basically remember is most of these users are coming over from windows or macOS they’re normal users. They’re not going to research. They just want things to work the way they have if you can’t go into your application, set it up the same way you did on the other platforms. It’s basically a non-starter the biggest push for getting users to come over to the Linux side. Is valve and gaming. Nothing else is making users. Consider the platform as there’s no reason for it from their perspective.
So not having the one desired package is not like having your mobile phone without the phone application.
That’s my two cents on the issue I think yes it needs to happen eventually but currently it’s just something that’s not viable. Linux is divided so unless fedora can get every other package managing distribution to change on this nothing will change.
Even KaOS has a guide to get the 32 bit packages to get steam running.
This doesn’t answer the question of what to do with the growing maintenance burden of keeping i686 support alive. I know there are (few) use cases where 32-bit compatibility is still important, but at what cost can - and should - we keep supporting those? And for how long?
If that the case, bring that to the council, who creates a strategy pointing to grow the user base and where the major actor now moving people to Linux is a massive gamer
An interesting point of reference here is that the Universal Blue products are heavily Flatpak-oriented. But in Bazzite (their gaming-oriented product), they don’t use Flatpak Steam - they install rpm Steam into the base image.
Of the three, I think the COPR is a good place to start to gauge the following:
- What is actually needed to make the Steam application work
- What is actually needed to compile those tools
- What methods might be best for doing that ‘outside’ of Fedora base RPMs (aka instead of glibc.src building it, fedwin-glibc.src is what does it or if some other way is needed.)
Going from this and other threads, there are a lot of Fedora ‘users’ of Steam. However the work needed to keep Steam working is causing more work for people who aren’t interested in it. We (as the project) need to make some subset of them into active Fedora ‘contributors’ in order to keep Steam working. Researching what is needed is probably the first step they can do.
I definitely agree with you there and it definitely could work so I think what would be the optimal solution has just to say unless it’s 64bit you just won’t support it. As people looking at the distro can just get a clear image from the homepage far easier than finding out after and being upset due to time wasted.
Maybe there should be some sort of a poll to rather gauge what people are actually doing on the platform something that’s just sent to the desktop rather than it being in the form where you’re only likely to get fans of two extremes.
Yes, I know there is the flatpak, but as you can see from above, I don’t think that would resonate to well. So as I said, just flat out saying you don’t support it as far easier and people I feel will be able to make a competent choice with that information.
Again, that’s just my take no need to dive too deep into it. I think valuing your time as a maintainer as well as taking into consideration the users time when they are not tech savvy when changes happens that they don’t understand.
Yes.
We also need to understand what it would take to turn our infra into an opt-in for i686 stance.
If I had had the time and technical ability to stand behind a proposal with my own sweat equity, I would counter wth a proposal to switch our infra to opt-in for i686 so we could on a first cut of this shake loose all the i686 builds that don’t have a compelling/maintained use case. And build a process by which maintainers affirm interest and ability in supporting i686 for specific packages.
If we took that approach could we drop like 90% of i686 builds? Could we onboard quality maintainers with a self-interest to take care of the remaining ones?
But since I dont have the time nor ability to stand behind that proposal, the best i can do is inception it as a least disruptive way forward.
The one alternative solution that was put forward here that sounds most as if it could work like that is doing something similar than what ELN does - i.e. listen to Fedora builds and rebuild stuff in a separate build target.
ELN also has dependency resolution mechanisms in place to make sure all dependencies that are needed to keep this set of packages buildable and installable, which is what exactly what we’d need for i686 packages too.
However, ELN works because it can fall back on rawhide builds in case something hasn’t been (re)built for ELN yet, but doing something ELN-like for i686 wouldn’t have this fallback (since the Fedora builds wouldn’t have i686 builds any longer) - so I don’t think it could avoid potential re-bootstrap problems ![]()
Can somone explain what’s meant by this in the change proposal?
I’m fairly new to Fedora so I think there’s some project-cultural nuance I’m missing here - I can’t tell whether it means “shrug, obviously this is aligned with the Fedora Strategy” or “shrug, obviously the Fedora Strategy will need to be adjusted in line with this” or something in between.
I think the solution should be dropping support for 32-bit packages that aren’t needed for compatibility. If 90% of 32-bit packages were to disappear overnight, most users probably wouldn’t notice. The ones people are concerned with in this thread for Steam and gaming are really impactful to the Fedora userbase, however.
As much as I’d like this change to happen, it’s too soon. This change would kill off projects like Bazzite entirely right as Fedora is starting to make major headway in the gaming space. Neal Gompa already pointed out basic use cases that would be broken even if someone built the packages Steam itself needs to function.
It’s also causing irreparable damage to Fedora from a PR standpoint. I have been inundated all day with people sharing news articles and being genuinely concerned Steam is gong to stop working on their Fedora/Bazzite machines. I would argue not only should this change be rejected, the proposal should be rescinded to limit further damage to Fedora as a project.d
Perhaps open a separate one to talk about changing build architecture to build fewer 32-bit packages?
That would break the games I play on a daily basis, so I really don’t want this to happen.
I think he’s speaking to the same concerns behind the proposal.
It takes people who care about the i686 builds to maintain them.
If the automation tooling are producing them as artifacts simply because it ‘can’ that’s not the same as maintaining them. Just because we can do a thing, doesn’t mean we should.
We need to be honest here about resources and workload. If theses are important to continue building, people with a self-interest to keep them operable will need to step forward and maintain them.
Right now we have a system that says package maintainer you must maintain the i686 builds even if you have no personal interest in them. Its probably time for that to change.
That does not necessarily mean we turn off i686 builds for everything. It may mean that we make sure there is an opportunity for people who are capable and interested in maintaining the i686 builds have the space to take on that work for the specific packages that are needed for the use case they actually care about.
Look here’s the deal, I doubt anyone actually cares about i686 at this point in the scope of Fedora. We don’t actually support the 32bit hardware any longer as a target. Maintaining these binaries is just drudge work to keep other software people do care about working. It’s not ideal, its not fun…It is what it is.
Let’s find a way to reduce the drudge work as far as we can, and lets create a space for the people who still find it necessary to do for the hopefully small subset of fedora libraries that need to be provided as i686. But what we can’t do is continue to require everyone to maintain i686 across the full project space as a shared commitment.
Said another way, we need to have an “orphaned i686 build” process that is similar in intent to the “orphaned package” process we have now. We need contributors who will take responsibility for maintaining the i686 builds specifically. But we also need a way to communicate that we need those contributors and point them to the opportunity to get invovled as other people step back from the work.
Just a user chiming in, I’m a big fan of flatpak for most things, but unfortunately steam flatpak lacks the functionality that packaged version has. Most notably VR hardware support.
I realize this is the niche of a niche but it needs to be said.
As a final user, the proposed Copr solution sounds reasonable enough, but, again, as a naive user, I have no clue about the work it would involve.
VR can in the Steam flatpak. The issue is that from within the flatpak sandbox, the SteamVR tries to run
sudo setcap CAP_SYS_NICE+eip ~/.var/app/com.valvesoftware.Steam/data/Steam/steamapps/common/SteamVR/bin/linux64/vrcompositor-launcher
but it can’t use sudo. However, running the command from outside the flatpak sandbox allows SteamVR to work.
The solution would be for Valve to update the SteamVR installer to work around the flatpak sandbox or ask the user to run the command manually.