Code review please for bug : 14929 - liborchestrator hits assert during creation of new partition for installation http://defect.opensolaris.org/bz/show_bug.cgi?id=14929
Webrev at : http://cr.opensolaris.org/~mattman/bug-14929/ With change for bug 13993, the GUI when collapsing down primary partitions does not remove shuffle primaary partitions upwards leaving all blank ones at the end of the 1st four pinfo elements. This was done as fdisk expects primary partitions to exist in the same pinfo location as the partition_id, and thus we can preserve partition id's correctly. There was still some logic in disk_parts.c that depended on the old way, which was the cause of this crash. Whilst investigating this bug I discovered a few other minor ones which I will fix with this putback aswell. disk_parts.c : - get_next_used_partition() - Change to start cycling through partitions after the current one. - om_validate_and_resize_disk_partitions() - Update debug messages to make more readable and include printing of partition_order structure element - When determing first partition in disk layout, it assumed for primaries this partition would be in index 0 if pinfo array, if not it must be an extended partition. This logic is incorrect, as the first disk partition for primaries could be in any of the 1st 4 pinfo slots. Check should be i < FD_NUMPART. installation-disk-screen.c : - restore_unused_partitions() - Add some extra debug output calls. - collapse_partitions() - libtd when initially reporting disk layout only populates primary elements in the pinfo array, it does not show unused chunks, these are calculated by the GUI and popped into the pinfo array by the GUI. As such these unused chunk elements should be cleared out completely before passing into om_validate_and_resize_disk_partitions(), and other used chunks should have their partition_order reduced correctly if unsued ones are removed. Added code here to perform this initialization.
