Snow Leopard on Unsupported PowerPC Macs

[B S Magnet]

Bad news and good news. I finally got the video card flashed. Took me hours but when i wanted to try the card in my G5 it turned out some PCI cards won't fit into an PCI-X slot. So i'm not 100% sure it flashed correctly because i couldn't try.

Anyway i am on the hunt for some old macs so maybe i find some and i can test it. Will let you know if i know more.

It’s possible this is due not to PCI-X, but to the voltage requirement of your card, relative to what the PCI slot is rated to support.

This guide from CERN shows how a 5V-rated 32-bit PCI card will not fit into a 3.3V 64-bit PCI/-X slot (like what’s equipped in Power Mac G5s):

How to recognise a 3.3 Volt PCI slot?

 

[pc297]

Dear all,

I can confirm that the Geforce 7800 GTX 512 Mb version PCIe (flashed) is supported on 10A190; at least for CI (hardware accelerated) and QE (enabled). Still the usual artefacts though, as in no folder names appearing in icon view when opening a drive in the Finder. Thanks to voidRunner for the ImageIO fix though (post #187)

Thanks for the great work!

Cheers,

 

[Project Alice]

I attempted to boot the installation of 10A190 that boots successfully on my PowerMac G5, on my iMac G5 and it KP'd within about 10 seconds of the boot screen. I'm hoping I can get it to work, since the iMac has a PCIe based GPU it might actually be worth using.

 

[pc297]

Hi all,

So I tried to see how much of the 10A380 kexts are still ppc (I did expect it to be low since Apple had made it clear by the release of this final WWDC DP that ppc stuff would go since SL would not support it). I compared it to 10A190 and 10.6.8.

I used two sets of commands in S/L/E:

either

method 1)

find . -type f -perm +111 -exec lipo -info {} \; | grep ppc | wc -l

or

method 2)

for i in *.{kext,bundle,plugin,ppp}; do j=`ls $i/Contents/MacOS/`; k=`basename $j`; lipo -info $i/Contents/MacOS/$k; done | grep ppc | wc -l

(as the bsd find does not support -executable and thus also returns non binary files with -perm +111 :/)

Anyway, here are the results:

method 1)

PPC kexts:

10A380: 42
10A190: 392
10.6.8: 67

method 2)

PPC kexts:

10A380: 51
10A190: 201
10.6.8: 51

I had a look at some of the apps, and a lot of them seem to be i386-only (notably the finder).

Conclusion: 10A380 is basically as ppc-free as 10.6.8 and will never boot as is.

I do wonder though about slapping its kernel on 10A190, or importantly the one from 10A222 (10.0.0d3). I think someone tried this with the one from 10.6.8 a few pages back and it did not work though unfortunately.

In any case, I guess would the numbers of ppc kexts in 10A222 is somewhat lower than that of 10A190 given the partial booting. Determining which ones are gone and replacing those with their 10A190 counterparts might be key to booting 10A222. Alternatively, the source is present for several of those in Darwin >=10.0.0; compiling them might be the key to solving the differences in kext loading in higher SL beta versions (i.e. 10A261 onwards)

On that note, it is also worth noting that several of the 10.6.8 ppc kexts have (unsurprisingly) higher version numbers than their 10.5.8 counterparts (and obviously 10A190), I will compile a list of those and see if they work in 10A190

Finally, re the Finder not displaying icon names properly in icon mode, I will check whether the finder from 10A222 is universal, and if it is, try to replace the 10A190 with it (or with the one from 10A96). Or with the one from 10A286 which is fully Cocoa. Might work since voidRunner had good results with the 10.5.8 Finder.

Cheers,

pc297

 

[pc297]

Sorry re 10A222 just did a bit of reading, as mentioned in post #86, #132 and https://www.insanelymac.com/forum/topic/144893-all-signs-point-to-no-ppc-in-106/, most executables and extensions in 10A222 including the Finder are Intel only; I just took a look at the exact number of universal extensions using the commands above; the results are similar to 10A380, maybe a slightly higher number of universal kexts (method 1: 47, method 2: 61) but by far nowhere near 10A190

So transplanting the missing kexts from 10A190 will not be enough, it should be the case for the missing executables too. A simpler solution might entail transplanting the 10.0.0d3 kernel, universal extensions and apps from 10A222 to 10A190 as mentioned in post #120

Thus compiling the latest versions of the open source kexts and kernel as well as using unique closed source bits of 10A190, updated universal kexts in subsequent releases (collecting the closed source ones all the way up to 10.6.8) as well as apps is the way to go - using the ones for which dependencies are still satisfied. Which, for the 10.8.0 kernel, is a safe bet, if it can even be made to work fully as a standalone ppc installation (afaik PureDarwin is struggling to this day to get 10.8.0 even as a standalone Intel Darwin distro, but mostly because they have to get rid of everything that calls closed-source software, mostly links to CoreServices and ApplicationServices)

Does anyone know if the 10A190 Finder is definitely the last one to be universal? Is it unreasonable to expect that it might have been compiled by chance to ppc and x86 in a later release as a one-off (e.g. like the 10.6.2 outlier, experimental 64-bit drivers for the GMA950)?

And did anyone try the 10A96 Finder in 10A190 to improve stability (would try this today anyway)?

Cheers

pc297

 

[pc297]

Most A190 x86 only binaries appear to be x86 specific and nothing to worry about, but AppleWWANSupport1 and BootCache seems like it would nice, Extensions.mkext is a file generated at boot if there's a mismatch between the kexts present and the kext cache mkext file, something like kextcache from the installer environment should suffice.

The source for BootCache is available for Darwin 10.0.0 but not AppleWWANSupport1. I will have a go at compiling it, unless anyone tried this?

 

[dermeister]

Has somebody thought of doing a binary file delta "report" on:

• 10A190 vs 10A432 (Last 10.6 PPC build vs 10.6 GM)

And then a comparison of

• 9L30 vs 10A432 (10.5.8 vs 10.6 GM)

...with the goal being to:

1. Systematically patch up A190 with 10.5.8 files that are same as 10.6 GM (release dates Aug 9 vs Aug 28 2009, so they are likely the same)
2. Have a list of soft differences that can be ported over from A432 (non executable assets)
3. Have a list of x86 executables that are different between A190 and A432. This would indicate where a PPC rebuild of an executable to match the differences would be required.

My two cents.

 

[Macbookprodude]

So, can someone tell me if this Snow Leopard is now ready to replace Leopard ?

 

[AtaruBarreau]

So, can someone tell me if this Snow Leopard is now ready to replace Leopard ?

No absolutely not sadly. It's the same as last year basically.

 

[pc297]

Sadly no luck using the 10A96 or even the 10.5.8 Finder in 10A190... The Finder simply does not show up when copying the entire app to CoreServices. Must be that it's part Cocoa in 10A190 :/ Thus the 10.5.8 Finder likely only works in 10A96 (I think voidRunner used the 10.5.8 Finder with success on 10A96). So I guess if the community is to use 10A190 as a basis to build upon rather than 10A96, we should try to understand what is not being drawn when it comes to icon names in icon view with the 10A190 Finder

 

[Macbookprodude]

Seems like we won't be able to fully use Snow Leopard as replacement for Leopard. Damn Apple for making PPC not useful too soon.

 

[MacPro2006VBox]

The best that I see is the use of a96 instead of a190, due to the fact that it has the most probable chance of coming out on top in terms of usability. The fact that it’s old enough to use the leopard finder without a fuss makes it a clear target for the project.

 

[Macbookprodude]

There is no way a quad G5 couldn’t handle snow leopard.

 

[MacPro2006VBox]

There is no way a quad G5 couldn’t handle snow leopard.

AGP G5’s are the only ones screwed over.

 

[Macbookprodude]

AGP G5’s are the only ones screwed over.

Exactly, only dual core machines and quad good for Snow Leopard.

 

[pc297]

Exactly, only dual core machines and quad good for Snow Leopard.

Isn't the 5200 FX PCI fully supported? (even though that's kind of a downgrade from AGP)?

 

[MacPro2006VBox]

Isn't the 5200 FX PCI fully supported? (even though that's kind of a downgrade from AGP)?

Yes that particular configuration is supported for some reason. I don’t know why the AGP one is not though. Maybe Apple was cutting out AGP support and was only supporting PCI/PCIE... a weird move that doesn’t make much sense.

 

[markker2992]

I attempted to boot the installation of 10A190 that boots successfully on my PowerMac G5, on my iMac G5 and it KP'd within about 10 seconds of the boot screen. I'm hoping I can get it to work, since the iMac has a PCIe based GPU it might actually be worth using.

Any luck with the iMac G5? Recently discovered this page and not sure how much I can help but I just came in possession of an iMac G5 iSight which is supposed to use the PCIe bus for the gpu to my knowledge.

 

[Project Alice]

Any luck with the iMac G5? Recently discovered this page and not sure how much I can help but I just came in possession of an iMac G5 iSight which is supposed to use the PCIe bus for the gpu to my knowledge.

I've been preoccupied with other old Mac projects recently. I will get around to it eventually.. No harm in you trying it out though.
It might work with a mostly clean install.

 

[Macbookprodude]

Can i get a status update ? Is QE and Acceleration now supported on G4s? I really want to install Snow on my PB G4 and move from Leopard to it.

 

[B S Magnet]

Can i get a status update ? Is QE and Acceleration now supported on G4s? I really want to install Snow on my PB G4 and move from Leopard to it.

If you aren’t seeing updates here from other volunteering efforts beyond your own, then the answer to your second question is probably “no” — with noted exception, of course, for G4s equipped with nVidia video GPUs.

 

[ChrisCharman]

Has somebody thought of doing a binary file delta "report" on:

• 10A190 vs 10A432 (Last 10.6 PPC build vs 10.6 GM)

And then a comparison of

• 9L30 vs 10A432 (10.5.8 vs 10.6 GM)

...with the goal being to:

1. Systematically patch up A190 with 10.5.8 files that are same as 10.6 GM (release dates Aug 9 vs Aug 28 2009, so they are likely the same)
2. Have a list of soft differences that can be ported over from A432 (non executable assets)
3. Have a list of x86 executables that are different between A190 and A432. This would indicate where a PPC rebuild of an executable to match the differences would be required.

My two cents.

I have considered doing this but am unaware of tools available for the platform that can do this, do you have any suggestions? I know there are many Linux open source options.

 

[Macbookprodude]

If you aren’t seeing updates here from other volunteering efforts beyond your own, then the answer to your second question is probably “no” — with noted exception, of course, for G4s equipped with nVidia video GPUs.

Ok, my PB G4 Titanium 1Ghz has Ati Radeon 9000.

 

[B S Magnet]

I have considered doing this but am unaware of tools available for the platform that can do this, do you have any suggestions? I know there are many Linux open source options.

Would a script involving diff be helpful in doing this?

 

[ChrisCharman]

Would a script involving diff be helpful in doing this?

I think we will need to use several tools, first to list all files that have changed and then to modify each file individually.
The highlighting of changed/updated files would be of great benefit - if this can be done using a diff script that would be an easily workable solution. Hex editors are capable of editing the files but I don’t fancy manually looking through all of the files in hexadecimal format and then trying to work out what it all means, as interesting as it would be i really don’t have the time spare at present.

Does anybody know of an open source decompiler that will produce readable code that may work for this project? I remember reading something about an MIT decompiler a while back but i’ll have to look into it further.

I have found a script on Github written by ‘jjazzboss’ that should display side-by-side differences, highlighted in colour, for binary files which could be useful for the files that don’t have open source code available i.e private apps and frameworks should anyone with experience reading an editing hex want to have a bash at backporting code changes: Colorbindiff.

With the initial list of differences established clearly between seeds and the GM, in terms of file changes and file structure changes, it would certainly be much easier to identify and recompile the rest of the binaries needed.