Adam Williamson
Adam Williamson
@pcdubs @travier @jmarrero
Yeah, I think it's an environmental thing; for FCOS it's sensible since the target environment is one where we know fallback path boot works and is commonly relied upon. But...
I don't think we really need to *not do* this in some environments (unless Colin disagrees), it shouldn't cause any harm at all in cloud environments. It's just not really...
So, hum. I don't think bootupd actually has the behaviour that argument is meant to produce. Looking at the way the EFI bootloader classes work, when that argument is set,...
I'm not sure you could manage "don't install a bootloader at all" that way, though. (edit: of course...we probably can't manage that *at all* with an ostree/container-based deployment, which will...
Filed https://bugzilla.redhat.com/show_bug.cgi?id=2268505 and proposed as a Beta blocker for the IoT case; it may not really be serious enough to make it a blocker, but I figured we could discuss...
So looking into how bootupd *names* the entry, it seems that it calls its own `find_efi_vendordir()` function and names the entry for whatever that returns. So it is, I think,...
Yeah, so far as efibootmgr invocations go that kinda sounds sensible to me. But then should we try and move that code out somewhere for anaconda and bootupd to share?...
> I don't have a sense of how serious this problem is. Maybe what would be easiest is to add an opt-in knob to do what this PR is doing....
yup, I am now applying this patch to Fedora downstream. having it merged upstream and a new version released would be great.