Comment 7 for bug 1262712

Splitting up the differences a bit there seems to be some changes to the packaging:
- drop and confict iscsitarget source
- recommend iscsitarget-dkms instead of module
- Recommend linux-headers instead of linux-headers-2.6
[Above all look sensible]
- Additional possibility to specify daemon options in default
- Removed some module assistant stuff (guess related to not recommend *-module)
- Using debhelper functions in postinst

Changes to kernel module code look to handle changes for newer kernels mostly (only exception seems to be some code added as persist.c). This has some big impact on the compat patches. Though IMHO the risk here should be low as the dkms code does compile and load at least on Precise and Saucy kernel installs.

Userspace changes look to be conforming to the changes described in the differences in ChangeLog:
 o Fixed iSNS initialization issue, thanks to <email address hidden>
o The NIPQUAD macro has been removed, instead "%pI4" is used
o Converted semaphores to mutexes
o restore SCSI SN behaviour of IET < 1.4.20 (VMWare compatibility)
o byte swap fix in MD5 / SHA1 code
o Updated initd scripts to reflect PID file name change
o Removed stray RPM build dependency on openssl from spec file
o Fixed crypto API error handling (ERR_PTR instead of a NULL)
o Fixed small memory leak in volume args buffer on BSD
o Fixed max_cmd_sn not being set for the session
o Fixed issue where StatSN was being updated when a command was ignored
o Fixed connection accounting in ietd for session reinstatements
o Fixed MaxOutStandingR2T handling
o Stripped zone id off ipv6 addresses in SendTargets
o Time2Retain fixed at 0 since it isn't implemented
o Makefile respects LDFLAGS
o Fixed re-enabling listening after max connections was exceeded
o Fix NULL-pointer dereference due to race in UA handling
o Added variable for depmod
o Don't install kernel module executable
o Don't use bash for initd script

Given that backporting the complete package should be only bugfixes and causes the dkms module and userspace be in sync (whether compiled on 2.6.32 or later kernel) it seemed to be smaller risk than to try to pick only the minimal changes to allow compile. The way things are mixed in the compat patches would make that not simple.

The Time2Retain change affects the config file but by default that had been commented out. And from the comment it does not sound like anything else than the new 0 default never would have worked.