Re: [bug] mkpartfs

[Andrew Clausen]

On Sun, Aug 05, 2001 at 02:22:57AM +0200, Neal H Walfield wrote:
> Consider the following transcript:
> 
>       address@hidden:~/src/parted-1.5.3/build (0)$ sudo parted/parted \
>       > /dev/hd2
>       GNU Parted 1.5.3
>       EXPERIMENTAL RELEASE - DO NOT USE ON VALUABLE DATA!!!
> 
>       [ chop ]
> 
>       Using /dev/hd2
>       (parted) print
>       Disk geometry for /dev/hd2: 0.000-19092.937 megabytes
>       Disk label type: msdos
>       Minor    Start       End     Type      Filesystem  Flags
>       1          0.031    953.859  primary   linux-swap  
>       2        953.859   1907.718  primary   sun-ufs     
>       3       1907.719   2861.578  primary   FAT         
>       4       2861.578   9538.593  extended              
>       5       2861.609   3815.437  logical   ext2        
>       6       3815.468   4769.296  logical   ext2        
>       7       4769.328   5723.156  logical   ext2        
>       8       5723.187   6677.015  logical   ext2        
>       9       6677.046   7630.875  logical   ext2        
>       10      7630.906   8584.734  logical   ext2        
>       11      8584.765   9538.593  logical               
>       (parted) mkpart
>       Partition type?  logical? log
>       File system type? ext2? ext2
>       Start? 10000
>       End? 11000
>       Error: The new partition overlaps with another partition.

This is subtle.

10000 - 11000 is outside of the extended partition, so the constraint
solver finds the closest match, which is something like 9530 - 9538.
However, that overlaps with partition 11 (hence the error message).

It's not clear what the best thing to do is, here...

The solution to the constraint is SO far away from what the user
typed... maybe this should be caught somewhere...

But, PedConstraint isn't expressive enough for partition tables
(it needs disjunctions).  Therefore, the solving is the responsibility
of the partition table code (*_partition_align()), so such checking
can't be controlled by the user of the API.  Maybe I should
bite-the-bullet TM and make PedConstraint support disjunctions...

>       (parted) resize
>       Partition number? 4
>       Start? 2861.578
>       End? 19092
>       Error: Can't grow a partition onto used space.

This one is even more subtle, and really ugly.  (Suggestions how to fix?)
Try typing 2861.579 instead of 2861.578.

I think the best solution is to add a resize constraint to make it
keep it within free space either side of it, IFF it has logical
partitions.

>       (parted) quit
>       address@hidden:~/src/parted-1.5.3/build (0)$ exit
> 
> As you can see, I have a 20G hard drive.  The extended partition (as
> created by cfdisk under GNU/Linux), only extends until the end of the
> last logical partition.  Therefore, I cannot create additional logical
> partitions without extending the extended partition.

Right, that's how it works.

> However, Parted
> will not let me resize the extended partition because it overlaps other
> partitions (but, of course it does!).

The problem is, the number you typed overlaps with partition 3,
a primary partition.  (rounding crap!)

> Parted should always assume that
> the end of the extended partition is the end of the disk: all primary
> partitions _must_ start and end before the start of extended partition,

Wrong.

> thereby, making this is a valid assumption.

s/valid/invalid/ ;)

Andrew