Syslinux - Apr 2014 - Bootloader data in /boot vs package systems (and atomic updates)

On Tue, Apr 22, 2014 at 8:16 PM, Gene Cumm <gene.cumm at gmail.com>
wrote:
> 
> I think you mean the VBR since no Syslinux variant lives in the MBR
> and on Linux, 
Right, thanks.
> Does the installation of an updated package of GRUB2 trigger an
> installation of modules during install on Fedora
It used to, but no longer does.  There's some background here:
https://bugzilla.redhat.com/show_bug.cgi?id=735259
> The alternative is an additional script that would run the installer
> binary extlinux and copy *.c32 at the same time.
Yeah, I'm considering changing at least the Fedora extlinux package to 
work this way.
> This would also be a distinct departure from past behavior.
Right, that's why I wanted to get feedback here first.  I'm not sure if 
other distributions have encountered this.  Maybe it doesn't matter 
much for traditional package systems as they're not robust against 
interruption at all anyways.

> 
> > The alternative is an additional script that would run the installer
> > binary extlinux and copy *.c32 at the same time.
> 
 
FWIW and just as one example, ArchLinux has its own script, 
"syslinux-install_update -i -a -m".
-i : install SYSLINUX;
-a : mark partition as "active";
-m : install mbr boot code.

I have not checked whether currently the syslinux-install_update.sh 
script supports UEFI (probably not).
 https://gist.github.com/pyther/772138 

Regards,
Ady.

On Wed, Apr 23, 2014 at 4:49 PM, Ady <ady-sf at hotmail.com>
wrote:
> 
> FWIW and just as one example, ArchLinux has its own script
Looks like the canonical source is here:
https://projects.archlinux.org/svntogit/packages.git/tree/trunk/syslinux-install_update?h=packages/syslinux

Right.  Hmm.  The painful thing will be transitioning the existing 
package, as it would obviously break everything if the updated RPM just 
removed the extant files from /boot.

I'm debating a bit whether or not this is worthwhile to do, or if I 
should just focus more on GRUB, or if I should patch Anaconda+OSTree to 
work around it.  Maybe both of the latter two.

Anyways, definitely a useful data point that Arch keeps the data in 
/usr and copies it to /boot via a wrapper.