Am 01.08.2019 um 14:13 schrieb H. Nikolaus Schaller hns@goldelico.com:

Hi,

...

Am 01.08.2019 um 13:45 schrieb aTc atc@k-n-p.org:

I don't think it should be unmounting the /boot partition on the machine running the script there.

It caused some issues booting here, since I had upgraded the kernel after running makesd, but it didn't end up on my /boot partition...

http://projects.goldelico.com/p/gta04-makesd/source/tree/master/makesd#L681

I see.

I have to research why it is there. Maybe it is a historic workaround for some no longer relevant problem and was not interfering on systems where /boot is not mounted.

Ok, the commit that did introduce it has the description:

"our letux fstab may mount multiple boot partitions to /boot and hide the one we want to reflash"

I think it is this: unter certain conditions there may be two mounts active on /boot, one hiding the other. If the hidden one is the eMMC boot partition and the Pyra has been booted from SD card, it is not possible to unmount the eMMC boot partition (because it is hidden).

The idea was to actively umount /boot (twice) to make sure it can be mounted as /media/P1. This appears to be only be needed if one runs makesd on the Pyra to flash eMMC that way.

I am not aware that another device needs this. Maybe the BeagleBoneBlack.

Well, this is a really rare case where a better solution should exist. So I have reverted the patch and you can do a "makesd -u".

BR, Nikolaus