The release of (lib)parted 3.0 includes a change to the Application
Programing Interface - API. Most importantly, libparted 3.0 removes
many file system specific function calls, and hence the capabilities
provided by these functions. In order for GParted to compile and link
with libparted 3.0, this libparted functionality is lost.
Specifically, the functionality that is lost when GParted is compiled
and linked with libparted 3.0 is as follows:
- Loss of ability to grow and shrink FAT16 and FAT32 file systems
- Loss of ability to shrink HFS and HFS+ file systems
- Loss of ability to determine used and unused sectors in HFS and
HFS+ file systems
- Loss of ability to erase file system signatures on partition
create and format
It is hoped that other free software projects will include some or all
of the above lost functionality, which can then be added back to
GParted.
This commit includes a change in how FAT16 and FAT32 file systems are
moved. Specifically the move is now performed internally by GParted
when linked with libparted 3.0. The move functionality is provided by
libparted for prior libparted versions (e.g. less than 3.0).
This is the final enhancement in a series of commits that enable
GParted to compile with libparted version 3.0.
Closes Bug #651559 - Doesn't compile against parted 3.0
On disks with an unrecognized disk label, the end sector value
was saved as one sector larger than actual. This caused the
Partition Information dialog to report the end sector value and
the total number of sectors on the disk as one larger than the
actual value.
Closes Bug #649444 - Partition information reports wrong sector
count
I hit this performance bug when I cloneda 40G NTFS partition. The actual
copy was done in under 11 minutes. After that I was shocked to find that
gparted would spend over 12 minutes in cleanup_cursor, pegging a CPU
core. (On a quad core desktop...)
Simply replacing the ustring with std::string would reduce the time to
about 1.5 minutes. Still bad. Also, I didn't want to lose UTF8
awareness.
So I rewrote the algorithm in 'streaming mode'. This has the (potential)
drawback that locale conversions are done, but performs well and
- IMHO - is a lot more readable.
On a minor note: this implementation correctly handles backspaces at the
start of a line.
Change spawn_command_line_sync to spawn_sync to reduce the level of
quoting required when passing commands to the function. Since
spawn_command_line_sync uses the parent's environment, the replacement
call also uses the parent's environment.
This change permits unmounting of file systems mounted with an
apostrophe in the mount point name.
Ubuntu launchpad bug 151796 - Can't unmount partition with apostrophe
in label
https://bugs.launchpad.net/ubuntu/+source/gparted/+bug/151796
Usage of the kpartx package to create partition names has been
completely removed. Hence kpartx is no longer even an optionally used
package by gparted.
This change is related to the following bug report:
Ubuntu launchpad bug 719129 - [Natty] Gparted duplicates dmraid
partition devices
https://bugs.launchpad.net/ubuntu/+source/gparted/+bug/719129
The new configure option --enable-libparted-dmraid enables the native
use of /dev/mapper dmraid support in the libparted library.
If this option is not specified, or is disabled, then gparted code
will be used to support dmraid. The gparted code uses the old dmraid
partition naming scheme of always appending the partition number to
the device name. The letter 'p' is not inserted between the device
name and the partition number. The gparted code is particularly
useful when used with older versions of libparted that do not support
/dev/mapper dmraid.
Ubuntu launchpad bug 719129 - [Natty] Gparted duplicates dmraid
partition devices
https://bugs.launchpad.net/ubuntu/+source/gparted/+bug/719129
Added a display refresh to the event queue to address this
problem.
Previously, the display area of the visual disk seemed to
disappear when enough operations had been added to the queue so
that a scrollbar was required (about 4 operations).
If the three lines in HBoxOperations::load_operations method for
"//make scrollwindow focus on the last operation in the list"
were commented out then this also prevented the problem from
occurring.
Interestingly, if I changed the "set_cursor" method to
a "scroll_to_row" method, then this problem does not occur for an
additional 2 operations (about 6 operations total).
Ubuntu launchpad bug 583746 - Partition graph disappears after a
swap partition has been deleted
https://bugs.launchpad.net/ubuntu/+source/gparted/+bug/583746
Previously when creating a new partition on a device with 512
byte sectors with 7 MiB chosen and cylinder alignment, the
snap_to_cylinder logic would round up to 2 cylinders (~15.69
MiB), instead of the correct 1 cylinder (~7.84 MiB).
Re-work logic for copying a cylinder aligned partition to a new
MiB aligned partition. Previously this was partially fixed in
bug #626946 - Destination partition smaller than source partition
Closes Bug #639393 - Redux: Destination partition smaller than
source partition
Use mntent C library to read mount point files such as
/etc/fstab, /etc/mtab, and /proc/mounts.
Closes bug #629592 - Unable to find Mount Point if it contains a
Space Character
The reason for refactoring is to simplify the large GParted_Core
class, to help minimize disk reads, and to group the logic for
processing the file /proc/partitions into a single logical class.
This reason for refactoring is to simplify the large GParted_Core
class, to help minimize disk reads, and to group the logic for
processing the file /proc/partitions into a single logical class.
Enhance code so that all calls to the ped_partition_get_path
function in the libparted library go through one
GParted_Core::get_partition_path method.
This one method makes the appropriate adjustments to the dmraid
partition path name as required.
Closes bug #634553 - ntfsresize fails on RAID array
This problem would occur when using align to MiB for the
destination partitions.
Closes bug #628863 - gparted crashes applying move to right of
more than one logical partition
This new URL is actually an html redirect to the actual page
location. This permits us the flexibility to move the actual
page wherever is needed on the web site without the need to
change the URL location in the code.
Correct problem introduced in version 0.6.3 whereby move or copy
operations using MiB alignment resulted in the destination
partition being 1 MiB larger.
Closes bug #632478
GParted already has basic support only to recognize Btrfs volumes; if
configured, use our newer methods to handle them.
Also, put in place interface controls for all supported generic actions.
This adds initial handlers for Btrfs; only .create, .check and
.read_label are done for now, via external btrfs-tools.
Other methods are still only stubs.
Change string initialization to use direct assignment instead of
a class method in the GParted_Core class initialization.
Thanks go to Jan Claeys for pointing out this potential problem
and providing the following link:
https://bugzilla.gnome.org/show_bug.cgi?id=627724#c7
This link contains a comment by Matthias Clasen that says:
"In a class init function, don't do anything that causes
other threads to initialize classed types."
Prior to this enhancement creating a new partition aligned to MiB
after a partition aligned to cylinder would result in the creation of
a partition that was 1 MiB smaller in size than requested. In the
case of a 1 MiB new partition, a warning message would be displayed
indicating that "A partition cannot have a length of 0 sectors".
This problem is avoided by applying similar logic as was used to
address the following bug:
Bug #626946 - Destination partition smaller than source partition
Prior to this enhancement when copying and pasting a partition
after a cylinder aligned partition, the destination partition
would become smaller than the source partition if align to MiB
was selected.
Closes bug #626946 - Destination partition smaller than source
partition.
Newer versions (dmraid-1.0.0-rc16?) of dmraid default to always
creating partition names by inserting the letter 'p' between the
device name and the partition number. Since this is not
consistent with older versions of dmraid, add extra logic to
ensure that newer versions of dmraid do not insert this extra
letter 'p'.
Rename method get_udev_name to get_udev_dm_name and enhance to
retrieve the property DM_NAME.
Handle situation where parted, and hence libparted, has been
configured with --disable-device-mapper. In this situation on
some distros the device name could be /dev/dm-0. This results in
invalid partition names such as /dev/dm-0p1. Add logic to decode
the actual /dev/mapper name.
With this patch the commands required to support motherboard BIOS
RAID, also known as fake RAID, are dmraid and dmsetup.
If the kpartx command is available, it will be used to maintain
device entries that follow the naming convention used by kpartx.
Closes Ubuntu launchpad bug 554582 - GParted does not list dmraid
devices
https://bugs.launchpad.net/ubuntu/+source/gparted/+bug/554582
This patch along with a few preceding patches address the
following bug reports:
Closes bug #622217 - gparted live 0.6.0-1 incompatibility with
AMD RAID (SB750 southbridge)
Closes Ubuntu launchpad bug 600729 - Gparted not usable on
fakeraid arrays
https://bugs.launchpad.net/ubuntu/+source/gparted/+bug/600729
Prevent overwriting meta data (Extended Boot Rectors) for logical
partitions by temporarily increasing the size of the logical
partition to encompass all of the space involved in the move
operation. The libparted library will move the EBR as needed to
permit this to happen. After the move the logical partition is
set to the proper size.
This fixes bug #623630 - Move logical partition to right yields
invalid partition table on /dev/sda - wrong signature 0
GParted requires libparted version 2.2 or higher to support devices
with sector sizes > 512 bytes.
This commit marks the end of a series of enhancements required to
provide support for sector sizes > 512 bytes.
Do not show unallocated space of exactly 1 MiB in size if:
a) The space is at the front of the drive.
This space is reserved for the partition table and the
Master Boot Record.
or
b) The space immediately precedes a logical partition.
This space is reserved for the Extended Boot Record.
Quote from GNOME Human Interface Guidelines 2.2.1 on Drop-down lists:
"Selecting an item from a drop-down list should not affect the
values of any other controls. It may sensitize, insensitize, hide
or show other controls, however."
Make align to MiB the default setting instead of align to cylinder.
Migrate logic for alignment to cylinder into its own method
snap_to_cylinder, and place common logic in snap_to_alignment.
Add alignment checks for situations where space is needed for Master
Boot Record or Extended Boot Record.
Adjust ranges on spin buttons according to required boot record space.
Copy fix for off by one sector (#596552) from
Dialog_Partition_New::Get_New_Partition to
Dialog_Base_Partition::Get_New_Partition
Enhance resize / move logic for checking locations of nearby logical
partitions to not depend on the partition ordering.
Note: This commit does not include limiting graphic movement according
to required boot record space.
Also add signal handler to alignment menu to update file system
minimum size.
This enhancement is to prepare for adding a third alignment
option to align to MiB.
This enhancement was added because the current libparted library,
version 2.2, does not appear to detect file systems for sector
sizes other than 512 bytes.
With the removal of the 512 byte constants, such as MEBIBYTE, it
was possible to rename the _FACTOR constants back to BYTE
constants. The _FACTOR constants, such as MEBI_FACTOR, were a
temporary measure to help in the transition to support devices
with sector sizes > 512 bytes.
The device sector size is needed in all calculations that convert
between sectors and bytes. The device sector_size is included in
the partition object because this object is used to contain
operation information in addition to actual partitions and
unallocated space. A second option was considered to pass the
device object to many methods, but this was considered a much
larger task that might not provide significant gains in
maintainability.
Previously this value was calculated to end on a cylinder
boundary. On today's larger hard drives, which simply emulate
Cylinders/Heads/Sectors, some disk space might remain after the
last cylinder. This space would always be less than a full
cylinder in size. Now this space will be visible to the user.
Improved logic in cleanup_cursor method to minimize the number of
string erase operations performed.
Previously when the NTFS copy finished, GParted would take an
exceptionally long time to clean up the output. This was due to
the huge number of small string erase operations performed on a
extremely large string.
A patch to alleviate the "failure to inform kernel of partition
changes problem" (bug #604298) has been included in parted-2.2.
Hence when parted-2.2 or higher is used, the work around code in
GParted is not required.
According to parted documentation, only "msdos" and "dvh" disk
types (or partition table types) support extended partitions.
All other disk types support primary partitions only.
This is not an ideal solution because devices with sector sizes
greater than 512 bytes are silently ignored from a graphical interface
perspective. Only if the user starts gparted from the command line
will the "ignoring device" message be seen.
In my opinion this short term solution of ignoring these devices is
still better than a gparted crash which might also result in data
loss.
A bug report requesting support for devices with sector sizes > 512
bytes has been created.
https://bugzilla.gnome.org/show_bug.cgi?id=607165
Thanks go to Jan Claeys for the initial small patch.
Bug #604298 involves a problem that can occur when GParted uses
libparted to manipulate a partition table, but the
ped_disk_commit_to_os() function fails to inform the Linux kernel of
these changes to the partition table. This problem is intermittent in
nature and is difficult to reliably reproduce.
This change is an attempt at a work around. If the first call to
ped_disk_commit_to_os() fails, then we sleep for 1 second to allow the
system to settle before invoking the function call again. My testing
has shown that this greatly improves the chances of successfully
informing the kernel of partition changes.
Thanks go to François Dupoux and Steven Shiau for their suggestions
and help to resolve this problem.
This problem was introduced with GParted 0.4.7 when the "-a"
option was removed from the dosfsck command. The dosfsck command
is used to read the number of used sectors. Instead we will now
use the "-n" option which does not write any changes to the file
system, and allows non-interactive operation.
Remove commented code and boolean use-C-lang parameter for
Utils::num_to_str() method because this bug was fixed in
gcc-4.0.3. The GNU gcc compiler versions 4 and up are now
commonly in use. Also due to previous partial commenting and
removal of the code, the code was non-functional
Thanks goes to Jan Claeys for pointing out this clean up
opportunity.
When calling commit_to_os() with newer Linux kernels,
parted-1.9.0, and at least one partition mounted on the device,
the following error message is displayed:
The kernel was unable to re-read the partition table on
/dev/sda (Device or resource busy). This means Linux won't
know anything about the modifications you made until you
reboot. You should reboot your computer before doing
anything with /dev/sda.
If this message is not cleared then it will remain in the
libparted_message array and could potentially be displayed along
with an error message for a subsequent scanned disk device.
When libparted reports a problem with the partition table, capture
this message and attach it to the "unallocated" partition that spans
the entire disk.
Examples of these types of problems encountered are:
/dev/sda: unrecognised disk label
Can't have overlapping partitions.
The ability to paste into an existing partition was disabled in
v0.4.0. This ability has been re-enabled along with a warning
dialog that indicates the data in the existing partition will be
lost if the copy and paste operation is applied.
Restore copyright entries by original author to those of his last
known repository commit titled "released gparted-0.3.4 on
LarryT's request." on Feb 25, 2007.
Add my own copyright entries for files in which I changed source
code. Files in which I only made spelling changes do not have my
copyright entry added.
Auto detection of Linux software RAID devices was lost in GParted
0.3.8. This was because device scanning by the libparted device
call ped_device_probe_all() was replaced with custom code within
GParted to scan /proc/partitions for devices.
The reason the libparted call was replaced was due to a long
scanning problem with ped_device_probe_all(), a non-existent
physical floppy device, and a BIOS setting indicating a floppy
drive existed. See bug #351753:
https://bugzilla.gnome.org/show_bug.cgi?id=351753
When used with parted-2.0, GParted would display a warning
message indicating "kernel unable to re-read partition table" on
every device scan for each device with at least one partition
mounted. Device scans occur when GParted starts up, immediately
after a series of actions are applied, and upon manually invoked
device refreshes. This, I believe, is excessive notification to
the user and would be annoying.
Improve the feedback to the user when creating or pasting a
partition on a device with no partition table found.
Prior to this enhancement if a user tried to create or paste a
partition onto a device with no partition table found, GParted
would present the user with a Create Partition Table dialog.
When creating a new partition this could cause confusion for
users who were trying to format the disk. The user could
mistakenly believe the choice of an MS-DOS file system was being
presented.
Similarly on a paste operation it could be confusing to be
presented with a dialog to create a partition table without any
explanation of why a partition table was suddenly required in the
midst of the paste operation.
Neither the new partition creation, nor the partition paste
actions were performed after the Create Partition Table dialog
was displayed.
Removed the "-a" flag from the dosfsck command that is used to
determine the number of used sectors in FAT16 and FAT32 file systems.
Related to GParted bug #569921
In order to find a match between the device paths in mount files and
partitions as found by libparted, the mount device must be a real path
and not a symbolic link.
Closes GParted bug #582687
Excerpt from parted-1.9.0 release notes follows:
In libparted, the linux-swap "filesystem" types are now called
"linux-swap(v0)" and "linux-swap(v1)" rather than "linux-swap(old)"
and "linux-swap(new)" as in parted 1.8, or "linux-swap" as in older
versions; "old" and "new" generally make poor names, and v1 is the
only format supported by current Linux kernels. Aliases for all
previous names are available.
Previously the read_mountpoints_from_file() method only expected
a space between arguments in files such as /etc/fstab. Valid
delimiters are now space and tab.
Related to GParted bug #582687
To further minimize the chance of accidentally creating a new partition
table, the following steps have been taken:
1) Rename non-standard Create button to stock Apply button.
2) Remove extra warning about erasing entire disk device.
3) Update GParted Manual accordingly.
Closes GParted bug #576374
Perform direct update of Number of Hidden Sectors in NTFS boot record.
This change removes the need for /usr/bin/xxd and /bin/dd.
Related to GParted bug #574389
It appears that different dmraid versions handle specifying a
single dmraid device differently.
dmraid-1.0.0.rc14 uses a partial name only,
such as "isw" or "isw_cjbdddajhi".
dmriad-1.0.0.rc15 uses a full name only,
such as "isw_cjbdddajhi_Vol0".
To maintain compatibility with the broadest range of dmraid versions,
the ability to specify a single dmraid device is being removed.
Some distros appear to display /dev/dm-# device names with
libparted. Since this fails a pattern match with the dmraid
device name, check with udev to see if a pattern match is
possible with the name returned from udevinfo or udevadm info.
For example:
/dev/mapper/isw_cjbdddajhi_Vol0 is the device name
/dev/dm-0 is a symbolic link pointing to the above device name
Normally, the blkid command will detect the UUID for a file system.
In cases where blkid fails to detect the UUID and the vol_id command
is available, then try using the vol_id command.
Newer systems might not have udevsettle. It has been replaced by
"udevadm settle".
Thanks to Lorenzo Cogotti aka Micia for this knowledge, and for a
proposed solution.
Move crypt-luks detection prior to libparted file system detection.
If libparted file system detection is performed first, then encrypted
file systems such as ext3 will be incorrectly detected by libparted as
plain 'ext3' when in fact these file systems are encrypted.
The GIT version of libparted returns types of linux-swap(new) and linux-swap(old), previously only linux-swap was returned.
svn path=/trunk/; revision=1024
If translated language is relatively long, then split the "Resize/Move" toolbar button text into two lines, and do not display the "Undo" button.
svn path=/trunk/; revision=971
- Changed title Features to File System Support.
- Renamed button Refresh to Rescan Support.
- Created expander for legend and added narrative.
- Removed columns detect and read.
svn path=/trunk/; revision=966
Curtis Gedak