We need at least 2 HFS+ partitions, one of size ~1GB to be used as RecoveryHD partition, and the other of size >30GB to be used for installing the system. I already had such partitions, but anyway I deleted and recreated them with parted (gparted can be used as well).

We need to install first these tools:

apt install gparted hfsplus hfsutils hfsprogs

This turned out to be more difficult than I thought, because usually you cannot find anything related to Mac outside of Mac App Store (MAS), and in order to access MAS you need to have a running Mac OS and an apple account. After lots of searches and failures, I managed to find this page with direct download archives: https://7labs.io/tips-tricks/macos-high-sierra-direct-download.html

The part that I used was RecoveryHDMetaDmg.pkg which contains an archive with the content of the RecoveryHD partition.

The *.pkg archives can be opened with the command xar, which is not a part of the standard Debian linux, so I had to install it manually, according to the instructions on this page: https://www.oueta.com/linux/extract-pkg-and-mpkg-files-with-xar-on-linux/

apt install build-essential libxml2-dev libssl1.0-dev zlib1g-dev
wget https://storage.googleapis.com/google-code-archive-downloads/v2/code.google.com/xar/xar-1.5.2.tar.gz
tar -xfz xar-1.5.2.tar.gz
cd xar-1.5.2
./configure
make
make install

Then I extracted the archive with the command: xar -xvf RecoveryHDMetaDmg.pkg. From the extracted files we are interested only on RecoveryHDMeta.dmg, which has the content of the partition RecoveryHD.

First we need to convert the file RecoveryHDMeta.dmg to a disk image, then mount this image, and finally copy its content to the recovery partition:

1. Make sure that we have the needed tools:

   apt install dmg2img hfsplus hfsutils hfsprogs
   

2. Convert the dmg file to a disk image:

   dmg2img RecoveryHDMeta.dmg RecoveryHDMeta.img
   

3. Mount the hfs+ partition of the disk image:

   umount recovery_image
   rm -rf recovery_image
   mkdir -p recovery_image
   
   devloop=$(losetup --find --show ./RecoveryHDMeta.img)
   partprobe $devloop
   mount ${devloop}p1 recovery_image
   

4. Mount the partition RecoveryHD:

   umount recovery_hd
   rm -rf recovery_hd
   mkdir recovery_hd
   
   parted /dev/sda print
   parted /dev/sda print | grep hfs+
   mount /dev/sdaX recovery_hd
   

5. Restore the content of RecoveryHD from the image:

   rm -rf recovery_hd/*
   cp -a recovery_image/* recovery_hd/
   

6. Cleanup:

   umount recovery_hd/
   rmdir recovery_hd/
   umount recovery_image/
   rmdir recovery_image/
   losetup --detach /dev/loop4
   

After replacing the content of the RecoveryHD partition with the one that we downloaded from the internet we have to restart the computer and boot it from this partition. In Mac this is done by keeping pressed the [Alt] key as soon as it makes the startup sound, then selecting to boot from the recovery disk. It is important to have an internet connection during installation, since the installation image will be downloaded from the internet.

After the recovery system is started, it allows us to use several tools. One of them is Disk Util, which I used to delete the 30GB partition and create it again. This step may not be necessary, but I had messed a lot with the partition table and didn't know how to restore it in a suitable state. Deleting and recreating that partition ensures that it is in a state suitable for being used by the installer. This also has the drawback of erasing all your previous content, but in my case this was OK, because I had not used it much and had nothing important on it. If I had something important on it, then I would make sure to backup it first, by mounting this partion from the Linux system first (similar to what was done on the previous section).

After that, I started the Mac OS X Installer and selected the 30GB partition as the place where to install the new system. The installation takes several hours because it downloads the installation image from the internet.

Note: MacOS High Sierra by default formats the disk with the format APFS, instead of the traditional HFS+ format. For some reasons, the HFS+ format is sometimes more preferred than the new one. This tutorial shows how to start the installer and tell it to use the HFS+ format instead: https://malcont.net/2017/09/how-to-upgrade-macos-to-high-sierra-without-filesystem-change-hfs-to-apfs/

The Mac installer will overwrite the boot menu so that it boots directly to the Mac OS. We can fix this with a recovery USB stick, for example with the Debian installer stick. It can restore the boot menu so that we can boot to Linux again.

However, the Mac OS installer had also messed with the partition table, by creating one more partition. This had shifted the number of the rest of partitions by 1, for example /dev/sda5 became /dev/sda6 etc. For this reason Linux failed to boot properly. However this can be fixed easily from a rescue shell, by editing the file /etc/fstab and fixing partition numbers accordingly. Then I also reinstalled GRUB:

grub-install /dev/sda
update-grub

• Use Linux to install OSX from a DMG extracted to a partition - without a Mac DVD
• Install macOS High Sierra on Mac – Direct Download
• Extract .pkg and .mpkg files with xar on Linux
• Mount the HFS+ partition of a disk image
• https://malcont.net/2017/09/how-to-upgrade-macos-to-high-sierra-without-filesystem-change-hfs-to-apfs/