Re: [Bug-xorriso] Test Proposal for [bug #46716] Protective MBR partition is not marked as bootable

[Thomas Schmitt]

Hi,

xorriso wrote under my insufficient control:
> >    MBR partition      :   1   0x80  0xee            1        32803

Alexander E. Patrakov wrote:
> OK, this is the bug. You have a 0xee partition, that indicates that GPT

Yep. Looks like a xorriso bug.
(How could this slip through when i looked at it ?)

-----------------------------------------------------------------
The presence of HFS+ lets it expect GPT and thus causes type 0xee
for the first partition.
Else it would have put out 0x83 or 0xcd.
Poorly tested in other situations because grub-mkrescue is my
only HFS+ user and is eager to really produce GPT.
-----------------------------------------------------------------

I was too lazy to check again after i added the -hfsplus options
in the second mail.
I now tested that my first proposal would have worked with OVMF
and would then have failed if HFS+ had been produced in later runs.
In that first mail of today, the resulting partition type was 0x83.

Shall i count this as good luck or as bad luck ?


A pebkac error was to omit xorriso command -hfsplus "on"
before letting the option proposer work.

  $ xorriso -hfsplus on -indev minimal-grub2.iso \
            -report_el_torito as_mkisofs

correctly proposes

  -hfsplus-file-creator-type chrp tbxj 
'/System/Library/CoreServices/.disk_label'
  -hfs-bless-by intel_bootfile '/System/Library/CoreServices/boot.efi'

Please add these options if you make further tests with
option -hfsplus enabled.
(Patching for now: boot flag in partition 1, type 0x83 in partition 1.)


> Unfortunately, fdisk does not allow me to replicate
> your desired partition tables exactly. I did want to swap partitions, but
> couldn't.

After i saw your dd operations i thought that you could do it
that way. :))


> Here is the layout that I have tested:
> ...
> MBR partition      :   1   0x80  0x00            1        32803

There is no need for type 0x00 here. We'd only need it for a
MBR partition which EFI shall ignore as much as possible, if we
decide to violate the no-overlaps prescription.

But here we go the (middle) legacy path:
MBR partition type 0xef together with disjoint other MBR partitions
of arbitrary type. Just not 0xee.


If we can omit GPT then it would be convenient to let partition 1
start at LBA 0. So we'd get a mountable ISO partition. (udev and
blkid do strange things when the first partition is not mountable
but the base device is.)

I have to point out that Vladimir Serbinenko was always opposed
to partition start 0.
The option  --protective-msdos-label  had that name before he
told me the need for GPT.

If my favorite (without that 0xee bug) becomes a sincere candidate
for an alternative grub-mkrescue result, we would have to ask
Vladimir for the reasons of his decision at BIOS-only times.


Have a nice day :)

Thomas