> On May 3, 2019, at 7:35 AM, Frank Scheiner <frank.schei...@web.de> wrote:
>
> That's true for all code, if the interface you use changes, you have to
> adapt your code.
The difference is where the breakage happens. If it were C code, for example,
an API break would actually cause the dependencies failing to build meaning you
would see it on the development side and the broken stuff would never get
shipped to the customer.
For scripted code, the breakage happens on the customer’s side which is what we
just saw. Furthermore, scripts can break through all kind of subtle things like
text encoding etc. Writing good bash scripts is hard.
> It could be discussed if there's a better interface
> available to get the needed information (partition index), but I
> couldn't find one that also provides a relation between partition
> information (bootable, method newworld, designated mount point) and
> index. If someone has an idea I want to hear it, my updated code is on [1].
>
> [1]: https://salsa.debian.org/installer-team/grub-installer/merge_requests/2
>
>> Btw, are there any news regarding tests with grub-ofpathname?
>
> Only for my 11,2 type G5 so far and there it doesn't work:
> ```
> root@powermac-g5:~# ofpath /dev/sda2
> /ht@0,f2000000/pci@9/k2-sata-root@c/@0/@0:2
> ## This path works...
>
> root@powermac-g5:~# grub-ofpathname /dev/sda2
> /ht@0,f2000000/pci@9/k2-sata-root@c/@0:2
> ## ...this does not
Okay, thanks. I’ll have a look at the GRUB upstream code for grub-ofpathname
then and report back.
> root@powermac-g5:~# nvsetenv boot-device
> /ht@0,f2000000/pci@9/k2-sata-root@c/@0:2,\grub
> ## Booting with this setting doesn't work.
> ```
Gotcha.
Adrian
