M1 Monterey (betas) fork

[donaldguy]

I suspect everyone who cares knows already from comments in #923 - but FYI, I am/have been operating a fork (building on the M1 work of @xorpse et al) that you can find at:

https://github.com/donaldguy/yabai

Administrivia

• Detailed build/Install instructions can be found below: M1 Monterey (betas) fork #1042 (reply in thread)
• I am a partisan of the fic-style branch naming conventions proposed (probably jokingly) in https://twitter.com/tesseralis/status/1271197776886370305 ergo the trunk branch is canon

The Present

it (mostly) works on 12.1 beta 1 (21C5021h) 2 (21C5031d), beta 3 (21C5039b), and beta 4 (21C5045a) (which was binary compatible with beta 2 as far as Dock went) and 12.0.1 RC 2¹ (21A559)

I am only running & checking on M1. There is a second HEAD branch (appropriately) named intel-maybe-probably-not, YMMV.

I have upgraded to 12.1 beta(s) and canon will continue to track those as they are released and I get them.

If 12.0.2… is forthcoming @mak1A4 has asserted he will update (M1) offsets.

For either the 12.0.* or the intel case, if someone wants to tackle double checking / updating the SA offsets, I outlined my (fairly naive) binary diffing approach here

The Future

@koekeishiya has an M1 pro on the way² and will be targeting 12.0.1 on it when he can³– so my fork is intended to be temporary and may become archived.

It may also get repurposed to run ahead of the-future for developer betas, which I am in the deeply-questionable habit of installing shortly after their release.

That said, per request, Issues are open on the repo now and I will do my best to review PRs.

Also, feel free (unless/until @koekeishiya says otherwise) to also talk in here.

Footnotes

1. I believe which was promoted to final 12.0.1? (at least the Dock.app was the same?) ↩

2. https://github.com/koekeishiya/yabai/issues/725#issuecomment-949484389 ↩

3. https://github.com/koekeishiya/yabai/issues/923#issuecomment-956198083 ↩

[jclusso]

@donaldguy appreciate your efforts. Could you add the proper procedure for using your fork. I'm struggling to figure out how to specify to brew to install your version.

[joshyork]

Thanks for putting this together. I'm on an M1 and Monterey (12.1 beta) and can confirm yabai is fully working for me again! I thought I was already following these steps during previous unsuccessful attempts. I'm not sure if upgrading to the beta (12.1 vs 12.0.1) was the difference maker or perhaps I was fudging one of the steps previously. Either way huge thanks to everyone involved for getting this working!

[jclusso]

@donaldguy thanks for that. I'm wondering how this works. In the past and even when it was broken I was able to get it partially working without disabling SIP. After I set it up can I re-enable SIP? Without it, I can't use some apps because they require it. Prior to Monterey this worked fine and I was able to use most of Yabai without an issue.

[neilsabde]

@donaldguy I followed all the steps and yabai starts up properly but when I open a new window in chrome, it is completely transparent and then I need to force quit it. Any idea on why this happens?

[neilsabde]

I'm using xorpse's fork and I have replaced the binary with the one in yabai/bin/yabai

[AStaroverov]

Thank you, everything is working!

[donaldguy]

I kinda doubt I can summon via mention into a discussion unbidden, but I'll try:

@pschwartz ! In re #923 (comment):

I am using skhd with Yabai. I have compiled the latest from source and tried donaldguy/yabai's latest release. I am running a MacBook Air with the M1 and didn't read issues before upgrading to Monterey.

Yabai is working and Skid is. But none of the space focus commands provide any return or work. I can move a windows between spaces, but doing a yabai -m space --focus 1 fails to move my current focus. The --check-sa has a return one of 0.

Any thoughts?

the one release on my fork is not up to date, it was just a one off bone-throw for the person who asked for one for Nix reasons.

Try building yourself per instructions above

---

... I'm a little squeamish encouraging people to install a pre-built binary of what like ... already behaves more than a little like a rootkit, even though I know I have no intention of abusing it.

Trusting Trust, moral hazard and all that.

but if someone wants to PR a GitHub Actions workflow for build and release, I'd consider merging it.

[joshtwo]

I also can't use focus space. Everything else works fine, even creating new workspaces.

[joshtwo]

This is what the error log shows:

could not retrieve window details.
sh: line 0: [: =: unary operator expected
could not retrieve window details.
sh: line 0: [: =: unary operator expected
could not retrieve window details.
sh: line 0: [: =: unary operator expected
could not retrieve window details.
sh: line 0: [: =: unary operator expected
could not retrieve window details.
sh: line 0: [: =: unary operator expected

[joshtwo]

Nevermind, I was loading the main repo yabai instead of the fork. I didn't notice until now.

[donaldguy]

For the record those error messages suggests that the immediate problem was that you were interpreting with sh a file (likely ~/.yabairc) designed for bash or zsh ... I think

Likely using ==s where a stricter POSIX test (aka [) wanted to see =s

or something like that. (== is only technically valid in [[ tests, but I think bash tends to be permissive about it)

[ppanagiotis]

Did someone tried this fork with 12.1 and intel cpu? Thank you!

[ai212983]

although it must be old xcode, here @donaldguy describes how to fix it.

[ai212983]

Works on my machine (tm), here's instructions

[LordSushiPhoenix]

Also seems to work on my intel mac, aside from the yabai -m space --focus N command, which gives no output.

[donaldguy]

For the record, -m space --focus does seem to work for me on M1

Again, no one has sussed out the Intel SA injection points - which you would need for this to work, I believe.

The instructions for how to do so (how I do so on M1 anyway) are here: #923 (comment)

PRs welcome

[donaldguy]

It's also possible, if the assembled versions of the functions are unchanged from Big Sur, that all someone needs to do is add the == 12 condition to the existing branch in

https://github.com/donaldguy/yabai/blob/866869158f853995c5c42c78f1e40b46e66dc700/src/osax/x86_64/payload.m#L204

etc

[donaldguy]

Made the single change for 12.1 beta 2: donaldguy@8668691

doesn't seem like much touched Dock

[joshtwo]

I often get crashes when the computer sleeps. Is anyone else experiencing this? I've got a crash log in case that helps.

[donaldguy]

It's a little useless without any debug symbols honestly.

If you are set up with Feedback Assistant, Apple will sometimes offer you (or at least me 😅 ) the chance to submit a bug report (which you don't actually need to do—I assume they probably don't want anyway - since they'd rather people not futz around like this—but initially say you're gonna), and then while the sysdiagnose stuff is running you can do a Quick Look on the .ips file (which you can reveal in Finder from the submit form, then press spacebar on) and see (and copy/paste) a crash report with symbols loaded (presumably after the fact?)

I've never been super clear on how/if you can force this. But that would at least tell us where in SkyLight its going down

I am not having this issue (nor do I think most people are), so I would try futzing around with your config and see if you can narrow it down. My guess is its maybe borders related (that's the area that was generally buggy enough that I've just given up for the time being)

EDIT: but its almost certainly not gonna be an option ever on a stable build like 21A344 where the Feedback Assistant isn't even part of the install.

[donaldguy]

(but it could also be an M1 Max issue? I am on a lowly M1: Original Recipe)

[donaldguy]

Also though, that logs says you are on build 21A344 (12.0.0) which I do not have offsets for (and may or may not have ever - as it seems like the betas branched from the MBP pro flashing builds early, going by the numbering)

Consider upgrading to 12.0.1 (21A559) which I believe (cause e.g. Wikipedia says so) went out on the public release channel? and these offsets should work on donaldguy@00bf5e4 - still there at HEAD too)

Or make bad choices like meeeee and run 12.1 betas

[joshtwo]

I'll upgrade to 12.0.1, thanks a lot!

[donaldguy]

I ran the analysis / binary diff and the offsets from beta 2 (21C5031d) seem identical in beta 3 (21C5039b); I've updated the OP and repo description to reflect compatibility thereof

[donaldguy]

Ditto on beta 4 (21C5045a)

[donaldguy]

Have users of the fork checked out koekeishiya's new 4.0 prerelease?

I haven't yet but I'm mostly inclined to switch over to it - maybe bring over the xorpse-fork style SA injection if it seems more reliable

I may archive my fork entirely; but I'm open to keeping it going if there's a reason.

Any thoughts, anybody?

[ghost]

Hi, my Mac version is 12. 1. I follow your Youtube video and download yabai successfully! I'm deeply appreciated your effort.

However, I encounter the problem that when I use make command. I cannot complie the extract.sh file with some invaild instruction error. Firstly, I think it might be the assembly syntax difference bewteen intel and AT&T. After STFW, I didn't find a way to use a different syntax or even figure out the principle. But finally, I solve it through switch my ARCH from x86_64 to arm64 by

$env /usr/bin/arch -arm64 /bin/zsh

And then, I compiled all the files smoothly. As I find some people with same problem in youtube comment, I just put my solution here and hopefully it cuold help others.

[donaldguy]

I archived the repo since 4.0 has been released