Why does Anaconda overrides user decisions?

[Chris Murphy]

On Fri, Jan 23, 2015 at 10:54 PM, Tim <ignored_mailbox at yahoo.com.au> wrote:
> On Fri, 2015-01-23 at 13:23 -0700, Chris Murphy wrote:
>> Using the 4 partitions as primary is bad practice because it prevents
>> additional partitions for no good reason. There isn't a negative to
>> having extended partitions, GRUB can even boot from a /boot partition
>> on an extended partition.
>
> Until you come to repartitioning a disc when you want to keep some of
> the stuff on it.  I've had discs where I wanted to remove the extended
> partition, and keep the first normal partition, then re-use the rest of
> the disc.  It refused to delete the extended partition because it
> insisted that there was a virtual partition inside it, but there wasn't.

When you say "it refused" what's it? Sounds like a bug though. EBRs
are really simple things, there's nothing magical about them, but an
EBR defines both its own partition start-end, and points to the next
EBR. So it's possible an EBR points to a corrupt EBR and that triggers
the problem. And it might even be a sort of mis-feature designed to
keep the user from inadvertently deleting something in such an
ambiguous state. Anyway, it's kinda hard to know what's going on here
without an actual example: partition data and what tool is rejecting
the modification.


> It's a horrid scheme, only surpassed in evilness by the bastard LVM.

OK well, it at least solves the non-contiguous regions problem of
partitions by presenting those regions as if they were contiguous. So
if you're going to reject MBR, EBR and LVM, you're left with GPT.
While that has redundant and checksummed partition data, it doesn't
solve the discontiguous space issue. And you're at the whim of your
firmware, whether it'll tolerate GPT without face planting.

-- 
Chris Murphy