[Gptfdisk-general] unexpected results from `sgdisk --largest-new`

[andrew b. <abe...@ar...>]

hello -

i'm seeing some unexpected (to me) results when running `sgdisk --
largest-new` and wondering if it's a bug or a misunderstanding on my
part. testing with GPT fdisk (sgdisk) version 1.0.4.

the documentation states:
       -N, --largest-new=num
              Create a new partition that fills the largest available block of
              space on the disk. You can use the -a  (--set-alignment)  option
              to adjust the alignment, if desired. A num value of 0 causes the
              program to use the first available partition number.

i've attached output with details, but the behavior can be summarized
as:
   1. create a 500g logical volume
   2. run `sgdisk --largest-new=0`; this creates a 500.0 GiB partition
      starting at 2048 (as expected)
   3. extend the lv to 600g
   4. run `sgdisk --largest-new=0` again; this ignores the add'l 100g and
      instead creates a 1007.0 KiB partition starting at sector 34
   5. an additional `sgdisk --largest-new=0` creates the expected 100.0
      GiB partition in the added space

on a whim, i also tried `partprobe` between `lvextend` and the 2nd
`sgdisk` invocation.  same results.  also tested `--largest-new=2` with
no change.

from my understanding of the option documentation, the `--largest-new`
invocation after growing the lv should have consumed that space instead
of making the tiny partition at the beginning of the device.  does that
match others' views/experience?  as a guess, it might have something to
do with the "last usable sector" calculation, which appears unchanged
at 1048575966 the first time it prints the info for what it does see as
a 600.0 GiB lv...

thanks in advance for any feedback.  if i don't hear anything i'll
probably file a bug on ubuntu's launchpad.

	andy

-- 
andrew bezella <abe...@ar...>
internet archive