Syslinux - Jul 2014 - isohybrid: slint64-14.1.iso: unable to find mac efi image

> Hi,
> 
> Ady wrote:
> >  cdrtools.sourceforge.net/private/man/cdrtools-3.01axx/mkisofs.8.html 
> >  Mac  Boot entries for the Apple Mac platform.
> 
> Interesting feature. But the El Torito specs of Phoenix and
> IBM of 1994 state that "Mac" would be Platform Id 2.
>   http://download.intel.com/support/motherboards/desktop/sb/specscdrom.pdf
>   Page 9, Figure 2, Offset 1 of Validation Entry.
 
I don't understand why you say "*But* the El Torito...". I
don't see
a difference with what the mkisofs manual says.

Where I see some kind of source for potential confusion (at least for 
some users) is (quoting Thomas, but he is not the source of the 
potential confusion):
"This program expects to find the UEFI image as second boot image and 
the Mac image as third boot image. Both with Platform Id 0xef."

My understanding is that:
 0=x86 , 1=PowerPC , 2=Mac 
are all non-EFI systems, which leaves "0xef" for "every"
(U)EFI
system, whether such (U)EFI system is EFI 1.0, UEFI 2.x, a Mac-EFI 
system, or some mix/non-standard firmware.

So, users should be more specific (or clear) when they refer to a 
"Mac" system, as the firmware could be non-EFI, or (some variant of) 
(U)EFI.

Just as a remainder, "-eltorito-platform 0xEF" is equivalent to 
"-eltorito-platform efi" in mkisofs.

I would also suggest, at least **for test(s)**, to use the latest 
mkisofs (ATM, 3.01a24), and perhaps mixing it with Syslinux 
6.03-pre19 (or newer). Although... The scripts/receipts for using 
Syslinux 6.xx in Slackware would need changes, and I am certainly 
_not_ suggesting to actually update Slackware to use Syslinux 6.xx.

Regards,
Ady.

Hi,
> I don't understand why you say "*But* the El Torito...". I
don't see
> a difference with what the mkisofs manual says.
The manual announces "Mac" for Apple Mac.
The program obviously writes Platform Id 1 in this case.
The El Torito specs say "Platform ID 0 = 80x86 1=Power PC 2=Mac".
In the sum this does not match.
A bug in mkisofs would be the simplest explanation.

> My understanding is that: 0=x86 , 1=PowerPC , 2=Mac 
> are all non-EFI systems
Yes. EFI is younger than this specification.
The Platform Id 0xef was obviously introduced later.

> So, users should be more specific (or clear) when they refer to a 
> "Mac" system, as the firmware could be non-EFI, or (some variant
of)
> (U)EFI.
The El Torito Platform Id 2 refers to non-EFI Macs.
Possibly with MC68000 processors although at that time,
Macs were mostly equipped with Power PC.

The "--mac" of isohybrid probably refers to Intel Macs.

But all these fine details only matter at boot time.
Didier's problem is at isohybrid time.
And there it is fully explainable by the source code of
isohybrid.c. This code was obviously added to support mjg59's
Fedora ISOs.

Of course one could easily modify isohybrid.c to tolerate
platform ids other than 0xef.
This would make sense if indeed some computer boots from CD
only via platform id 1 (or 2). It could be tested without
isohybrid.
For the GPT and APM aspects, the El Torito platform id is of
no significance.


Nevertheless i advise Didier to simply try with 0xef rather
than Mac. At least the ISO will then have MBR, GPT and APM
entries for the file macboot.img.


Have a nice day :)

Thomas

> Hi,
> 
> > I don't understand why you say "*But* the El Torito...".
I don't see
> > a difference with what the mkisofs manual says.
> 
> The manual announces "Mac" for Apple Mac.
> The program obviously writes Platform Id 1 in this case.
> The El Torito specs say "Platform ID 0 = 80x86 1=Power PC 2=Mac".
> In the sum this does not match.
> A bug in mkisofs would be the simplest explanation.
 
OK, so the "But" was not about any difference between the El Torito 
specification vs. the mkisofs manual, but about the difference 
between those two vs. the actual result in Didier's case.

It might be interesting to test with mkisofs 3.01a24 and using (only) 
the "Mac" platform (and without any other, even without
"x86") and
see the result (on such "Mac" hardware).

Even without such "Mac" hardware at hand, it might still be helpful 
to test mkisofs 3.01a24 and then check the resulting ISO image. 
Perhaps Didier can update his current version of mkisofs and retry 
with his current scripts/receipts.

According to the subject of this email thread, we are talking about a 
"mac efi" anyway, so the platform ID should not be zero, nor 1, nor 
2, and the '-eltorito-alt-boot -no-emul-boot -eltorito-platform Mac' 
is being misused.

Regards,
Ady.

Hi,

On 29/07/2014 21:34, Thomas Schmitt wrote:
 > Nevertheless i advise Didier to simply try with 0xef rather than Mac. 
At least the ISO will then have MBR, GPT and APM entries for the file 
macboot.img.

I did that, and also tried the xorriso command you posted.

I uploaded both ISO images on http://slint.fr/misc/MacBoot resp. renamed:
slint64-14.1-Thomas_mkisofs.iso
slint64-14.1-Thomas_xorriso.iso

I still don't know if one can boot a Mac off an USB stick with these
images dd-ed (and won't be able to check before tomorrow, hoping that
one of my computer repairmen neighbors will accept to try, as I own no
Mac)

At least I did check that there's no apparent regression, as I could
boot using each of these two ISO images as virtual CDs on a vmplayer
VM (BIOS as well as EFI firmware), off an USB stick on a vmplayer
VM with an EFI firmware, and off an USB key on my physical Thinkpad T61.

So so far, so good... Expect news tomorrow hopefully for a test on a Mac.

Meanwhile, if one of you happens to have a Mac at hand and can try
booting it of an USB stick with one of these ISOs that would be great of 
course!

Thanks and have a night day,

Didier