[parted-devel] libparted and partition system id

[Jim Meyering]

Petr Uzel wrote:

> On Mon, Jun 27, 2011 at 11:07:52PM +0200, Jim Meyering wrote:
>> Petr Uzel wrote:
>>
>> > On Sat, Jun 25, 2011 at 11:08:44PM -0700, ben wrote:
>> >> Using libparted is there a way to set the system ID of the partition, for
>> >> instance 0x83 for linux, without using libparted to actualy write
>> >> a filesystem
>> >> to the partition?
>> >
>> > As far as I know, this is not possible with vanilla parted. However,
>> > you finally forced me to post here our SUSE patch that 'abuses'
>> > partition flags in (lib)parted for setting partition types. Attaching
>> > version that applies to parted-2.4.
>> >
>> >
>> > Jim, if I clean (at least the testsuite has to be adapted) the patch,
>> > would you be willing to accept it?
>>
>> Hi Petr,
>>
>> This looks useful.
>>
>> However, shouldn't any type-setting scheme try to be useful with GPT
>> partitions in addition to MSDOS ones?  I'd be happy to be able to set
>> the types of only those two partition table types, for now.  Of course,
>> just *how* to specify a GPT partition type is not as clear cut as it is
>> for MSDOS.  Do you want to always require a full UUID?  Short-cuts for
>> common values?  Once you provide interfaces to set these things, it's only
>> logical to expect an interface to get/print the values.  When thinking
>> about setting and especially printing arbitrary GPT partition type UUIDs,
>> I begin to think it'd be best not to try to shoe-horn this functionality
>> into the "flags".
>
> Hi Jim, Karel and all,
>
> You are right - it should be possible to at least set partition type
> for GPT and msdos (and maybe others - Karel mentioned mac).
>
> For GPT, I also agree that it should be possible to specify both full
> UUID as well as shortcuts for some of the most commonly used values.
>
> As for printing the type - do you think that extending 'parted print'
> with type information is sufficient or would you prefer some other
> way (command) of printing it? IMHO 'parted print' is fine.

Printing more information via the "print" sub-command would be nice,
but would cause trouble for any program that parses that output.
One possibility would be to output the additional information only
when some new command-line option is specified.

Another possibility, which I prefer, is to add a new command to print
a table with this new per-partition data.  Later we may want to print
(and even set) disk and partition "unique UUIDs" (aka GUIDs), too.

> And yes, using flags does not seem like a best idea, even if it worked
> :) For setting the partition type, I can imagine new command, 'parted
> type' which would call partition-table-type specific function to query
> the user and then set the type.
>
> I'd like to work on this (as well as on FS-agnostic move and resize
> commands we discussed recently). Right now I'm busy with SLES stuff
> and will have vacation soon (no computers!! :) ), but I won't forget
> this.