macOS 10.15 Catalina on Unsupported Macs

[0403979]

Very rarely.
[doublepost=1560083439][/doublepost]2) Replacing OpenGL+OpenCL + GPUSupport + Appkit : does reach stage2 apple logo, but then loop on an ACM:: functionIdontremember
(I've searched inside the SLE and it should be the AppleCredentialManager.kext)

So what if we replace AppleCredentialManager.kext?
[doublepost=1560087385][/doublepost]

I tried this and as you said, Catalina is working in the background with a blank screen and interestingly, when I connect my MB through the mini display port, the external screen just flashes and I can see (somehow) the Catalina wallpaper (?)

If we can ssh into Catalina during this and check logs and kexts that are loaded, that might get us somewhere.

 

[jackluke]

So what if we replace AppleCredentialManager.kext?
[doublepost=1560087385][/doublepost]
If we can ssh into Catalina during this and check logs and kexts that are loaded, that might get us somewhere.

Haven't tried yet, but consider that I've tested Appkit without the CoreDisplay and Skylight replacement.

Because I replace one by one to check the behaviour in verbose mode, when I notice too pid crashes I simply avoid to continue and force a power-off, but maybe adding more frameworks to the Appkit replacement, checking with "otool -L" could resolve those other pid crashes.
[doublepost=1560087910][/doublepost]

I tried this and as you said, Catalina is working in the background with a blank screen and interestingly, when I connect my MB through the mini display port, the external screen just flashes and I can see (somehow) the Catalina wallpaper (?)

Yes, maybe with a vnc session could work too, but we need some @ASentientBot opinions too before continue.

Because he discovered the correlation between Mojave 10.14.5 and CoreDisplay.framework.

 

[0403979]

Haven't tried yet, but consider that I've tested Appkit without the CoreDisplay and Skylight replacement.

Because I replace one by one to check the behaviour in verbose mode, when I notice too pid crashes I simply avoid to continue and force a power-off, but maybe adding more frameworks to the Appkit replacement, checking with "otool -L" could resolve those other pid crashes.
[doublepost=1560087910][/doublepost]

Yes, maybe with a vnc session could work too, but we need some @ASentientBot opinions too before continue.

Because he discovered the correlation between Mojave 10.14.5 and CoreDisplay.framework.

No offense to @ASentientBot but I think that discovery was based on the name and a bit of luck.

 

[Larsvonhier]

Maybe Siri requires acceleration?

nope, runs on MB4,1 without accel with Mojave...
[doublepost=1560088303][/doublepost]

@Larsvonhier your SiriUI .3 Mojave downgrade patch doesn't work more on Catalina, Siri appears for few seconds then auto-closes.

I've tried to downgrade around all the others Siri* privateframeworks but nothing.

How did exactly debugged it ?

Because apparently I can't notice a real Siri.app crash on Catalina.

I filtered all console log outputs according to the suspects I tried to verify. That pointed me in the right direction...

 

[jackluke]

No offense to @ASentientBot but I think that discovery was based on the name and a bit of luck.

Many scientific discoveries that changed the world were based on a good dose of luck supported by knowledge.

 

[jbiser361]

Many scientific discoveries that changed the world were based on a good dose of luck supported by knowledge.

that’s true

 

[pkouame]

nope, runs on MB4,1 without accel with Mojave...
[doublepost=1560088303][/doublepost]
I filtered all console log outputs according to the suspects I tried to verify. That pointed me in the right direction...

Yes. Fully debugging the acceleration issue requires turning on all of the internal framework logging capabilities and inspecting those ad nauseum. Beyond kernel verbose logging, Console and system logs should be our friends.

Now that I have my Cat neatly cloned to internal disk, it is faster and better disposed for some experiments.

Turning on all available framework logs and debugging options is my track. There are some new variables if you inspect the framework code (like in OpenGL and AppKit). Understanding why it fails to accelerate and downsizes to framebuffer should provide some better direction, but this may take more time.

Many scientific discoveries that changed the world were based on a good dose of luck supported by knowledge.

Someone way smarter than me once said genius is 1% talent and 99% sweat. You increase the luck factor by dogged determination which is the discoverer's special attribute...

 

[TimothyR734]

cMP3,1 managed to reach the welcome/choose_language screen, however the usb keyboard and usb mouse (both connected) were not recognized or responsive so had to shut down. any ideas?

you might have to use the usblegacyinjector.kext

 

[jackluke]

Yes. Fully debugging the acceleration issue requires turning on all of the internal framework logging capabilities and inspecting those ad nauseum. Beyond kernel verbose logging, Console and system logs should be our friends.

Now that I have my Cat neatly cloned to internal disk, it is faster and better disposed for some experiments.

Turning on all available framework logs and debugging options is my track. There are some new variables if you inspect the framework code (like in OpenGL and AppKit). Understanding why it fails to accelerate and downsizes to framebuffer should provide some better direction, but this may take more time.


Someone way smarter than me once said genius is 1% talent and 99% sweat. You increase the luck factor by dogged determination which is the discoverer's special attribute...

Just tried this: OpenGL+OpenCL + GPUSupport + CoreDisplay + Appkit + Skylight from Mojave (10.14.3) and same behavior they reach stage2 followed by a blank LCD, but it's not turned off, it's just backlighted in black color.

I've also tried alternating as you suggested using .4 Mojave CoreDisplay the one with VirtualDisplays functions, but same result.

Premised that Catalina still boot with its stock frameworks, until now the only way to boot using some legacy frameworks is downgrading OpenCL + OpenGL (and maybe also GPUSupport).

edit:
Only think I just noticed I used the AGC and GPUWrangler downgraded for the DualGPUs machine, should try with stock Catalinas. But @Minh_Ton tried too without them, so I don't think they are the issue, since I use them on 10.14.5 and acceleration works with no issues.

 

[hvds]

Many scientific discoveries that changed the world were based on a good dose of luck supported by knowledge.

...and extra knowledge gained by experiments - as you do!

 

[pkouame]

I was poking around using my external Catalina from another macOS to retrieve and put some apps, and noticed that:

Third party apps are stored here : /Volumes/APFS Data/Device/Applications/
macOS core apps are stored here: /Volumes/APFS Catalina System Label/System/Applications/

Again, I think the tricky APFS Volume to handle to create a Catalina USB Installer for unsupported Mac, is the "APFS System" rather than "APFS Data".

A good high-level summary of of the Catalina filesystem situation is from Bombich (of CCC):

"...In the Finder you'll only see one volume that represents your startup disk and it will appear as if everything is on that single volume. In reality, the startup disk that you see in the Finder is the read-only system volume and doesn't have any of your data on it; the "Data" volume is separate and hidden. CCC will have to find that hidden volume (don't worry, I know where to look 🙂 and back up its contents alongside the system volume. The current version of CCC can't do that, of course, it knows nothing about APFS volume groups. We're adding that functionality right now, though, and we'll be hard at work over the summer on some new documentation that will help explain everything that you're seeing (or not seeing!) on the newOS."

Apple is really trying to secure user's data and lock down the system stuff - what a pain (for us any way)

 

[jackluke]

A good high-level summary of of the Catalina filesystem situation is from Bombich (of CCC):

"...In the Finder you'll only see one volume that represents your startup disk and it will appear as if everything is on that single volume. In reality, the startup disk that you see in the Finder is the read-only system volume and doesn't have any of your data on it; the "Data" volume is separate and hidden. CCC will have to find that hidden volume (don't worry, I know where to look 🙂 and back up its contents alongside the system volume. The current version of CCC can't do that, of course, it knows nothing about APFS volume groups. We're adding that functionality right now, though, and we'll be hard at work over the summer on some new documentation that will help explain everything that you're seeing (or not seeing!) on the newOS."

Apple is really trying to secure user's data and lock down the system stuff - what a pain (for us any way)

Don't know, I had the impression they added, apart SIP, another layer of protection to the APFS_SystemVolume, I mean from Catalina single user mode I am unable to use the "kextcache" due to a new security framework (?) and unable to execute "fsck -fy" that asks me to put the container in read-only that is already, since I don't launched "mount -uw /".

 

[pkouame]

Just tried this: OpenGL+OpenCL + GPUSupport + CoreDisplay + Appkit + Skylight from Mojave (10.14.3) and same behavior they reach stage2 followed by a blank LCD, but it's not turned off, it's just backlighted in black color.

I've also tried alternating as you suggested using .4 Mojave CoreDisplay the one with VirtualDisplays functions, but same result.

Premised that Catalina still boot with its stock frameworks, until now the only way to boot using some legacy frameworks is downgrading OpenCL + OpenGL (and maybe also GPUSupport).

edit:
Only think I just noticed I used the AGC and GPUWrangler downgraded for the DualGPUs machine, should try with stock Catalinas. But @Minh_Ton tried too without them, so I don't think they are the issue, since I use them on 10.14.5 and acceleration works with no issues.

I believe what your are noticing is that the retro-patches are sufficient to get through certain boot stages. This makes sense since at that point the new Cat display handling and new functions are not entirely necessary.
When it comes to actually starting a Cat UI Session (login etc.) the old frameworks are simply not sufficient. This time around retro-patching may require even more of the Core/System frameworks, but at that point you can no longer call this Catalina - right? 😉 That's the fine line we are walking....
[doublepost=1560095258][/doublepost]

Don't know, I had the impression they added, apart SIP, another layer of protection to the APFS_SystemVolume, I mean from Catalina single user mode I am unable to use the "kextcache" due to a new security framework (?) and unable to execute "fsck -fy" that asks me to put the container in read-only that is already, since I don't launched "mount -uw /".

Have you tried recovery mode?

 

[jackluke]

I believe what your are noticing is that the retro-patches are sufficient to get through certain boot stages. This makes sense since at that point the new Cat display handling and new functions are not entirely necessary.
When it comes to actually starting a Cat UI Session (login etc.) the old frameworks are simply not sufficient. This time around retro-patching may require even more of the Core/System frameworks, but at that point you can no longer call this Catalina - right? 😉 That's the fine line we are walking....

Yes who knows maybe CoreGraphics or Carbon or Quartz or IOAccelerator frameworks, there is even a OpenAL framework, too many.

edit: yes, Catalina Recovery mode does work fine.

 

[Jonathan50]

you might have to use the usblegacyinjector.kext

thank you for your input. ticked that one but kbd & mouse still unresponsive.

 

[jackluke]

@ASentientBot premised your IOHIDFamily fix works fine to speed up the boot process, I noticed that also in Mojave as Catalina there was this folder /System/Library/HIDPlugins but in Catalina apart they added a bunch of new frameworks, in particular I noticed they added in PrivateFrameworks these HID.framework and HIDAnalytics.framework .

And also /SLF/IOSurface has a dedicated all its own framework, this also in Mojave and HighSierra.

About non-metal acceleration, I was thinking about IOAccelerator privateframework could this be involved in some way.

Or should we try to downgrade also IOSurface.framework ? Since the kernel panic for legacy Video Framebuffer on Catalina occurred from there in some way.

 

[Starplayr]

Someone way smarter than me once said genius is 1% talent and 99% sweat. You increase the luck factor by dogged determination which is the discoverer's special attribute...

When I am doing something new and don't have the a right technical know how, I come up with creative ways of solving the problem and 90% of the time it works. May not be how someone who trained for years in Engineering but there is something to solving problems creatively and the best part is your left brain over time will catch up and understand why your off the wall solution even works in the first place.

 

[flygbuss]

A good high-level summary of of the Catalina filesystem situation is from Bombich (of CCC):

"...In the Finder you'll only see one volume that represents your startup disk and it will appear as if everything is on that single volume. In reality, the startup disk that you see in the Finder is the read-only system volume and doesn't have any of your data on it; the "Data" volume is separate and hidden. CCC will have to find that hidden volume (don't worry, I know where to look 🙂 and back up its contents alongside the system volume. The current version of CCC can't do that, of course, it knows nothing about APFS volume groups. We're adding that functionality right now, though, and we'll be hard at work over the summer on some new documentation that will help explain everything that you're seeing (or not seeing!) on the newOS."

Apple is really trying to secure user's data and lock down the system stuff - what a pain (for us any way)

I wonder how time machine will handle it. Did anyone try it yet?

 

[Starplayr]

One thing I notice with block level copies is they are so good, even the UUIDs of the APFS volumes are the same.

The problem with that is if you clone to another disk and both disks exists, because of the UUIDs you can use the other disk, even if you bless the other clone manually it seems to just boot the first one.

Is there a way to change the UUIDs of APFS files, even if it is something random afters it's been imaged?

The other other solution for my dd images will be just keep disk image snap shows for that one specific volume and for other disks do a fresh install for those volumes and that will keep the UUIDs separate. A bit of a pain, but it's a workaround that will work.
[doublepost=1560096679][/doublepost]

I wonder how time machine will handle it. Did anyone try it yet?

Yea that one virtual volume thing is a bit of a gotcha. I have tested CCC's latest beta for Catalina and it's not ready for prime time (yet).

In the meantime, a front end for making and restoring 'dd' images could be written. The only gotcha is duplicate UUIDs and could pose a problem when both disks are live on a system (the original and clone has the same UUIDs from a block level copy). Now if you are just cloning and restoring from lets say a subsystem or another install to the same volume that won't be a problem and with dd you can keep snap shots if your drive to a large volume like an external 4TB drive and dd when used properly is ultra fast.

pkouame -> I saw a great hike in performance with rdisk.

this did an awesome job:

sudo dd if=/Volumes/OtherStuff/catOS-A.img.dmg of=/dev/rdisk8 bs=1m

I might start using a combination of dd and rsync. Use dd to create a install of vanilla system and use rsync to take file data from an established system to the new clone to prevent having to setup certain things from another volume while keeping the UUIDs unique as the new clone will be made from new install.

I still want to visit hacking the latest OSInstall.mpkg and have it be both edited and signed and that way could do installs. That is until a patcher tool is created. For me it will be a good experiment as I don't have a valid OSInstall for my 3,1 (yet).

 

[pkouame]

One thing I notice with block level copies is they are so good, even the UUIDs of the APFS volumes are the same.

The problem with that is if you clone to another disk and both disks exists, because of the UUIDs you can use the other disk, even if you bless the other clone manually it seems to just boot the first one.

Is there a way to change the UUIDs of APFS files, even if it is something random afters it's been imaged?

The other other solution for my dd images will be just keep disk image snap shows for that one specific volume and for other disks do a fresh install for those volumes and that will keep the UUIDs separate. A bit of a pain, but it's a workaround that will work.
[doublepost=1560096679][/doublepost]

Yea that one virtual volume thing is a bit of a gotcha. I have tested CCC's latest beta for Catalina and it's not ready for prime time (yet).

In the meantime, a front end for making and restoring 'dd' images could be written. The only gotcha is duplicate UUIDs and could pose a problem when both disks are live on a system (the original and clone has the same UUIDs from a block level copy). Now if you are just cloning and restoring from lets say a subsystem or another install to the same volume that won't be a problem and with dd you can keep snap shots if your drive to a large volume like an external 4TB drive and dd when used properly is ultra fast.

pkouame -> I saw a great height in performance with rdisk.

this did an awesome job with rdisk:

sudo dd if=/Volumes/OtherStuff/catOS-A.img.dmg of=/dev/rdisk8 bs=1m

Do you mean beyond changing the UUID folder names in Preboot and Recovery?

 

[sylvioG5Quad]

macOS Catalina runs great and very smooth on MacPro5,1. No need to patch or upgrade.

You just need to do these 3 simple steps for the MacPro5,1 to run macOS Catalina:

(1) install macOS Catalina on a drive on a supported Mac

(2) boot your MacPro5,1 to recovery, and then in terminal, type

sudo nvram boot-args="-no_compat_check"

(3) transfer the drive with macOS Catalina installed to the MacPro5,1 and then boot. The MacPro5,1 will now boot up macOS Catalina.

this is not working......terminal fail....

 

[jackluke]

this is not working......terminal fail....

without "sudo".

 

[Starplayr]

This is interesting:

when you compress a dd image with Disk Utility and open the compressed image,

It mounts the Preboot volume, Recovery volume with access to its UUID folders.

And you can see the contents of those folders.

VM, Preboot, Recovery and the combined System volume all get mounted from the compressed disk image. I may be able to convert it to read / write and change the UUIDs and clone it again with different UUIDs with dd.

Talk about luck.

I don't know about the UUIDs of the system volume. but work a try changing the Preboot and Recovery to get those to be a unique pair the clone.

You can all access the Boot plist file on the preboot volume and BaseSystem.dmg
[doublepost=1560098172][/doublepost]

this is not working......terminal fail....

I run the sudo command from Mojave. As long as SIP is off it will work:

sudo nvram boot-args="-no_compat_check"

 

[pkouame]

One thing I notice with block level copies is they are so good, even the UUIDs of the APFS volumes are the same.

The problem with that is if you clone to another disk and both disks exists, because of the UUIDs you can use the other disk, even if you bless the other clone manually it seems to just boot the first one.

Is there a way to change the UUIDs of APFS files, even if it is something random afters it's been imaged?

The other other solution for my dd images will be just keep disk image snap shows for that one specific volume and for other disks do a fresh install for those volumes and that will keep the UUIDs separate. A bit of a pain, but it's a workaround that will work.
[doublepost=1560096679][/doublepost]

Yea that one virtual volume thing is a bit of a gotcha. I have tested CCC's latest beta for Catalina and it's not ready for prime time (yet).

In the meantime, a front end for making and restoring 'dd' images could be written. The only gotcha is duplicate UUIDs and could pose a problem when both disks are live on a system (the original and clone has the same UUIDs from a block level copy). Now if you are just cloning and restoring from lets say a subsystem or another install to the same volume that won't be a problem and with dd you can keep snap shots if your drive to a large volume like an external 4TB drive and dd when used properly is ultra fast.

pkouame -> I saw a great hike in performance with rdisk.

this did an awesome job:

sudo dd if=/Volumes/OtherStuff/catOS-A.img.dmg of=/dev/rdisk8 bs=1m

I might start using a combination of dd and rsync. Use dd to create a install of vanilla system and use rsync to take file data from an established system to the new clone to prevent having to setup certain things from another volume while keeping the UUIDs unique as the new clone will be made from new install.

I still want to visit hacking the latest OSInstall.mpkg and have it be both edited and signed and that way could do installs. That is until a patcher tool is created. For me it will be a good experiment as I don't have a valid OSInstall for my 3,1 (yet).

Yes using rdisk is so much faster! But do keep in mind that as an unbuffered transfer it is a little more risky. Especially if it's tunneling to another machine on the network. Using rdisk on image creation is also a time saver. I'm thinking a 32GB image should be a decent size to play with next time.

I used rsync when attempting my manual approach. I failed to create an APFS boot using it, but I don't think it had anything to do with rsync - still there are some reports that rsync has a few bugs and memory leaks. Also setting up rsync required an extra brew install step while straight dd is out of the box.

 

[Starplayr]

Yes using rdisk is so much faster! But do keep in mind that as an unbuffered transfer it is a little more risky. Especially if it's tunneling to another machine on the network. Using rdisk on image creation is also a time saver. I'm thinking a 32GB image should be a decent size to play with next time.

The clone for 32GB is so fast it works really well. The disk can be compressed down and save 20%.

I have not been able to do a restore from Disk Utility. Which for ease of use is a bummer. I am thinking about writing frontend for dd.

I used rsync when attempting my manual approach. I failed to create an APFS boot using it, but I don't think it had anything to do with rsync - still there are some reports that rsync has a few bugs and memory leaks. Also setting up rsync required an extra brew install step while straight dd is out of the box.

I am using 32 GB as well. 🙂 I don't have WiFi, Sound yet. boot can be slow if networking is not working. I need to catch on this thread for possible fixed for my MP3,1. Using Ethernet to a Mac mini over WiFI and USB Sound works, so I am not terribly worried about it.

For rsync or ditto, I was thinking about just cloning the Data volume to another Data volume of two working OS systems. basically migrate your data volume easier on fresh installs. (same users, etc).

I am working on a clone with the UUIDs edited to see if I can have a clone and original exist on the same system. Not sure if there is a checksum in the UUIDs, just changing one character. If that doesn't work, I will just format new disk as APFS and get its UUIDs and then clone to it using those UUIDs.
[doublepost=1560100522][/doublepost]@pkouame Oh for rsync and ditto, you will need to give terminal Full Disk Access to the terminal. It will ask you to relaunch Terminal, but a reboot may be required.

just did another restore:

30769414144 bytes transferred in 137.502351 secs (223773731 bytes/sec = 223.77 MB/sec! )

That is fast but it is going from a PCIe SSD to my SSD Hardware RAID in one slot. It does not run as full speed as a newer system would, but it's fast enough on this old rig.

Since Mojave it is hard finding good prices on Mac Pro 5,1s. I got this 3,1 for 90 bucks from work. I have another 3,1 that we use as a gaming PC and to flash 680 cards when needed. I only flash stuff for myself now. And latched onto this Galaxy 680 card which for a 2GB card, I swear its faster than any EVGA 680 I've run.