(In reply to pgnd _ from comment #4) > > [Note to self: verify all parts of your claim *before* posting!] Which "self" do you think this remark referred to? (IMHO) Obviously to me, as I wrongly claimed that impossibility. > snark is helpful how? Well, if you want to misunderstand me, you definitely will (anyhow, I'm no native speaker and do make mistakes (see above)). > > No -- definitely not! Not your command -- that's IMHO impossible. > > It's worked for well over a year. No, it did not! You could have used efibootmgr -c without any argument to achieve a comparable result. You specifically have *never* selected an EFI System Partition (ESP) with '--part-uuid'. > That it's, in your terminology, 'luck' doesn't change that fact. Really? Try placing an ESP on 'sdb', rearrange disks or add another host controller, which happens to shuffle "drive letters"... > And doesn't excuse the lack of either error or documentation. Hmm, that 0.6.0 didn't error out is indisputably unfortunate, but not retroactively fixable. > > a) no "parameter error" reported for misuse of '--part-uuid' -- it's > > simply ignored! > > there's no indication in "--help" that it's not valid. You could take the solitary reference in '--delete' as a hint! [...] > certainly reads as 'create for/using the selected variables' as viable, or > at the very least NOT proscribed. Well, that's your opinion. As I never use '-b' to select the number for a boot entry to create, I wrongly assumed, that everybody only ever would want to "select" an entry for deletion. > an error message, then, would be nice. Granted, but the error message in 14 is apparently also "not good enough". > as would a manpage Sorry, but again "no". The command you're looking for is rpm -V efibootmgr which would show, that the manpages have been removed/omitted from your installation. Does it? [...] > man efibootmgr > No manual entry for efibootmgr Not here on my system, but it wouldn't have helped either, as that SUSE-specific extension hasn't been documented in the manpage (yet). It was simply *not* intended for public consumption (but for installer use, namely to avoid stale boot entries). [...] > > b) your boot partition residing on '/dev/sda' with number '1', which > > matches the compiled-in default of 'efibootmgr'. > > no. "/boot" resides on RAID /dev/md0, on part 3 Aargh -- sorry -- I meant to refer to "ESP". 'efibootmgr' is never concerned with a '/boot'-partition. Unly the partition hosting '/boot/efi' is relevant to that program (specified with '-d -p'). [...] > Unclear. > > I used to clear the boot order ages ago, as grub2/Xen/UEFI/efibootmgr on > Opensuse had been creating a downright mess. That's since cleaned up, and I > simply write, not clear. Unclear as well -- I simply do not understand this and the next statement. > With kernel 4.[11-12], appeared as usual. Atm, with 4.13x, not. That's > being invesitgated separately. > > In any case, the primary point is the writing of the BootOrder with my > intent, rather than a depending on FW (most FWs are robust; some are clearly > horked) Every time you successfully create a boot entry with 'efibootmgr -c', it should recreate 'BootOrder', with that new entry being the first. If it can't there should be an error message. If you intentionally remove 'BootOrder' (with 'efibootmgr -O'), it would (IMHO) be useful to edit out the error message regarding that variable from a submission (to avoid unnecessary diversions). > "--disk ... --part ..." works fine. Good. And this is the only way forward. But I'll take your challenges into account, while upstreaming this '--delete' extension (together with a proper manpage-part), which is work-in-progress. BTW, it would be helpful, if you could close that issue #75, as upstream can not (yet) be of any assistance in this topic.