On Sat, Mar 29, 2014 at 01:51:46PM -0400, Phillip Susi wrote: > Loop labels were incorrectly identifying the device for the fictional > partition as $dev1 instead of just $dev. This caused other programs like > gparted to be confused, and caused parted to fail to identify the partition > as busy due to the fact that it was looking for the wrong device. Parted > also actually created the partition device so your raw fs on $dev gained an > alias as $dev1. Next, writing the label back to the disk clobbered the > filesystem there if it used the first sector. Several filesystems end up > using the first sector for 2048/4096 byte sectors even though they don't > for 512/1024 byte sectors. Finally, fat and ntfs boot sectors were being > detected as msdos labels. > --- > NEWS | 2 + > include/parted/device.in.h | 1 + > libparted/arch/linux.c | 44 +++++++++++++-------- > libparted/disk.c | 2 + > libparted/fs/ntfs/ntfs.c | 2 +- > libparted/labels/dos.c | 29 ++++++++++++++ > libparted/labels/loop.c | 42 +++++++------------- > partprobe/partprobe.c | 4 +- > tests/Makefile.am | 1 + > tests/t1102-loop-label.sh | 96 > ++++++++++++++++++++++++++++++++++++++++++++++ > 10 files changed, 174 insertions(+), 49 deletions(-) > create mode 100644 tests/t1102-loop-label.sh
This makes a whole pile of tests fail. I think I also previously commented that I wasn't seeing the problem as described -- I need to dig up that email and read it again. -- Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)