[U-Boot] cmd: part: unify syntax of uuid according to start/size subcommands

[Stephen Warren]

On 4/25/19 1:36 PM, Eugeniu Rosca wrote:
> Hi Roman,
...
> On Thu, Apr 25, 2019 at 02:18:22PM +0300, roman.stratiienko at globallogic.com wrote:
>> From: Roman Stratiienko <roman.stratiienko at globallogic.com>
>>
>> This allows retrieving uuid of the partition using it's name.
> 
> IMHO not seeing any real-life motivation (backed up by use-cases and,
> ideally, some commands/console output) in the patch description is the
> right recipe for getting no feedback from community. Fortunately, I am
> willing to put some time to fill this gap.
> 
> The story which I see behind the patch is that you are unhappy about
> not being able to pass the partition name in order to get the partition
> uuid. Currently, 'part' does require the partition index as input:
> 
>   => part uuid mmc 1:1
>   d117f98e-6f2c-d04b-a5b2-331a19f91cb2
>   => part uuid mmc 1:misc
>   ** Bad partition specification mmc 1:misc **
> 
> It looks to me that pretty much the same driving force guided:
>   - http://git.denx.de/?p=u-boot.git;a=commitdiff;h=36df616a2
>    ("cmd: part: Allow passing partition name to start and size")
>   - https://patchwork.ozlabs.org/patch/1044151/
>    ("[U-Boot,v3,1/7] cmd: part: Add 'number' sub-command")
> 
> So, the motivation is clear to me (and I share it!).
> 
> The problem which I see with this patch is that it changes the usage
> pattern of the 'part uuid' sub-command, which breaks current (mainline
> and potential out-of-tree) users of 'part uuid'. Below occurrences in
> u-boot/master will require an update if this patch is merged as-is:

Yes, I don't think you want to change the cmdline format for this 
command, or all existing use-cases will break. That's not just the 
scripts you mentioned, but also people's own scripts stored on their own 
boot media or U-Boot environments, or just their memory of how to run 
the commands.

Right now IIRC the following works:

part uuid mmc 1:1

Perhaps we can make the following work, where the partitionk ID 
parameter isn't a simple integer, or where that ID doesn't exist:

part uuid mmc 1:"partname"

That should be backwards-compatible, and a sane enough syntax.

BTW, it looks like the patch also re-orders a bunch of code while 
editing it. I guess this is to keep the code that handles all the 
sub-commands in alphabetical order, which is fine. However, such cleanup 
should be a separate patch, so that the patch which introduces 
new/changed behaviour /only/ does that, so it's clear what's going on.