On Mon, Jun 23, 2014 at 12:29:53PM +0200, Roman Neuhauser wrote:
hello,
Arvin mentioned in libstorage/pull/69 that there are "important and fundamental changes required in libstorage". i'd like to know what you (both Arvin and the team as a whole) plan in this regard?
I don't plan anything since the feature requests use all of my
time anyway. Years ago I started a to-do list that just keeps
growing. Some entries:
- continue real-world testsuite
- generate image in build-service
- more tests
- automatic runs
- move some stuff from GetPrimPartitons to libstorage (for kiwi)
- container and volume design starts to break down:
- md can be container or volume since kernel 2.6.28
- people want to use whole disk as volume with filesystem
- pv on whole disk cannot be encrypted
- XEN disks
- btrfs?
- snapshots?
proposal: split filesystem/encryption stuff from Volume into
new class Payload and derive Container and Volume from that new
class Payload
- use a graph instead of list of list
- separate parsers from logic
- for dmpart/multipath/mdpart there are *two* partition objects
for each partition!
- think about proposal rewrite in libstorage
- don't remember udev data in DmPartCo but use the one in Disk
instead - ugly (derive DmPartCo from Disk (as Dasd does it)?)
- detect mount also in inst-sys and remove AddSwapMp?
- unify commit messages, e.g. some contain size others not,
create vg is destructive while create sw raid isn't
- don't even look at disks used by BIOS RAID or multipath (see
bnc #460081, #436825 c36)
- avoid calling external grep
- add ronly_ptable for disk so that ronly has always same meaning
- maybe move ronly to Device
- use more data from /sys during probing
Regards,
Arvin
--
Arvin Schnell,