Changes for GPT fdisk version 0.8.1
diff --git a/NEWS b/NEWS
index 40953f8..7e62f33 100644
--- a/NEWS
+++ b/NEWS
@@ -1,16 +1,29 @@
-0.8.1 (?/??/2011):
+0.8.1 (10/1/2011):
------------------
+- Fixed bug that could cause FixParts to keep a partition's assignment
+ as logical when FixPart could not actually do so. This could happen
+ when there are no gaps between two logical partitions. Some partitioning
+ tools can create such configurations, but FixParts can't. Such
+ configurations are extremely rare. I've only encountered them when
+ logical partitions are out of order.
+
+- Added code to detect infinite loops of logical partitions when reading
+ MBR data. When detected, the program now stops reading after the first
+ loop, so no duplicates appear in the partition list.
+
+- Fixed bug in partition overlap detection in MBR code.
+
- Changed GPT reading code to use the size encoded in GPT headers to
determine how much of the header to use in computing a CRC, with the
restriction that the size be equal to or less than the disk's sector
size. This should work around problems with libefi in ZFS, which sets the
- header size to 512 rather than the standard 92. A caveat: If the disk's
- sector size is larger than the GPTHeader data structure size (512 bytes),
- then the rest of the sector's contents are ignored and replaced with 0
- values. This could produce false positives on CRC checks on disks with
- over-512-byte sector sizes if the header sector is padded with something
- other than 0 values.
+ header size to 512 rather than the more common 92. A caveat: If the
+ disk's sector size is larger than the GPTHeader data structure size (512
+ bytes), then the rest of the sector's contents are ignored and replaced
+ with 0 values. This could produce false positives on CRC checks on disks
+ with over-512-byte sector sizes if the header sector is padded with
+ something other than 0 values.
- Fixed bug in new (as of 0.8.0) check that main and backup partition
tables are identical on big-endian (PowerPC, etc.) hardware.