Re: parted script mode fails with unit option

[Jim Meyering]

Petr Uzel wrote:

> On Sun, Oct 09, 2011 at 10:30:59PM +0530, Keshav P R wrote:
>> Hi,
>>      Parted in script mode (latest git snapshot) fails to  run when "unit 
>> MiB"
>> is used (mainly for mkpart commands). I am planning to use similar commands 
>> in
>> Archlinux installer script. Please check the attached file 
>> test_parted_unit.sh
>
> Hi,
>
> If the user specifies position of start/end of the partition with
> units like MB, kB, cyl and similar, parted is allowed to adjust the
> specified start/end of the partition in some sensible range - IIRC
> this is always half size of the unit in both directions. See the note
> about IEC binary units in parted info page.
>
> However, this 'feature' is disabled for following units: sectors, MiB,
> GiB, and similar. Therefore, if you create the partition starting at
> 1 MiB and ending at 2 MiB, it will occupy sectors 1024-2048 (inclusive).
>
> Following request to create partition from 2MiB to 202MiB fails,
> because that would mean the starting sector (2048) would overlap with
> the last sector of the first partition and parted is not allowed to
> fix it.
>
> I have just submitted a patch that makes the message more informative
> by including exact partition geometry in sectors:
>
> ./parted/parted -s /dev/sdd unit MiB mkpart PART 2 3
> Error: You requested a partition from 2.00MiB to 3.00MiB (sectors 2048 - 
> 3072).
> The closest location we can manage is 2.00MiB to 3.00MiB (sectors 2049 - 3072)
>
>
> What might be even better solution is if the user specifies the _end_
> in IEC units, the partition created by parted would actually end one
> sector before.
>
> Advantages:
> - users would not run into this issue if using only IEC units
> - no unnecessary gaps between partitions and alignment constraints
>   still met (AFAIU we don't care if the partition end is aligned)
> - partitions created using IEC units would also have size divisible by
>   the unit (without the extra sector)
>
> Disadvantages:
> - yet another change of IEC units semantics
> - different semantics for start and end
> - more bytes to document it :)
> - ???
>
> What do you guys think?

I like it.

> If this sounds good, I will submit a patch.

Please do.