[Solved] NVIDIA Web Driver iBooks Bug

[vahids]

Nvidia Web Driver - 367.15.10.15f03 --> build 16B2657 is out but still not fixed the issue. :-/

 

[xoplague]

Nvidia Web Driver - 367.15.10.15f03 --> build 16B2657 is out but still not fixed the issue. :-/

Also having this issue. Guess this just means my iPad is finally useful again >;D

 

[vahids]

Also having this issue. Guess this just means my iPad is finally useful again >;D

But there there are not any Instruments App for my iPad :-/

 

[onx]

The same issue GTX970 with iBook and Xcode instruments. This bug didn't looks like something fatal because I even could use context menu in iBook App and it's not transparent.

 

[zakklol]

@rimkashox reading through the documentation for kernel extensions (kext), it seems you need a special signing certificate to make this work: https://developer.apple.com/library...l#//apple_ref/doc/uid/DTS40007919-CH1-TNTAG19

Removing the code signature won't work, as code signed apps that use library validation (such as Xcode, iBooks, etc) won't load unsigned or invalidly signed libraries.

This is really in NVIDIA's court. Hopefully they've just used the wrong cert.

It's more complicated than this, though. The things referenced in the error messages are NOT kexts, they are bundles loaded dynamically into processes that need them. I think it happens deep down in system frameworks depending on what you use.

So it's not a matter of using a cert that has kext signing abilities, as that's not relevant for these bundles.

I'm not sure this is actually 'fixable'. If metal and ogl drivers exist as bundles that are dlopen()'d into a process at runtime, then there's no way to have third party drivers work. They will fail for any application signed with the -o library option.

I have an app that I develop that makes extensive use of Core Animation and OpenGL. Works great on hackintoshes+Nvidia web drivers and real apple hardware. I did a quick experiment and changed my build process to sign with -o library and it renders the app completely broken on the hackintosh.


The rules for Library Validation are that the loaded library either be a 'system library' or have the same team identifier as the process doing the loading. Basically for system level stuff like metal/gl drivers, the only way to accomplish that is to have apple provide them; I see no mechanism for 3rd parties to generate signatures that would work in this scenario (nor should there be, really).

 

[zakklol]

One quick follow up to this: if you have some other non-nvidia metal/ogl capable card installed, instruments and iBooks will work just fine since CoreImage/Animation will favor other hardware backed renderers over the software one. In my case I had to have the proper inject intel+ig-platform-id setup in clover so my integrated graphics showed up properly in the system report. No need to have a monitor connected to the IGP.

Once that's done instruments and iBooks are working again. They will work even if you are displaying them on a monitor connected to the nvidia.

 

[sladej]

Anyone try the new 367.15.10.25f01 drivers with 10.12.2 to see if this issue is fixed?

 

[onx]

One quick follow up to this: if you have some other non-nvidia metal/ogl capable card installed, instruments and iBooks will work just fine since CoreImage/Animation will favor other hardware backed renderers over the software one. In my case I had to have the proper inject intel+ig-platform-id setup in clover so my integrated graphics showed up properly in the system report. No need to have a monitor connected to the IGP.

Once that's done instruments and iBooks are working again. They will work even if you are displaying them on a monitor connected to the nvidia.

Hi! You use on-board Intel graphics?

 

[carbonimax]

Anyone try the new 367.15.10.25f01 drivers with 10.12.2 to see if this issue is fixed?

Yes and it doesn't fix the issue.

 

[zakklol]

Hi! You use on-board Intel graphics?

Yes, but with no display connected to it.