[Pkg-zfsonlinux-devel] About downstream patches on debian packages from zfsonlinux.org repository.

[Turbo Fredriksson]

On Aug 29, 2016, at 8:12 AM, Petter Reinholdtsen wrote:

> [Brian Behlendorf]
>> But they can't be accepted upstream until they're reviewed and verified
>> to work properly on other major distributions.
> So, how do we get in touch with people able to test in other major
> distributions?

Because of the way 'libshare' works, this is a somewhat bogus requirement.
Libshare is a wrapper, nothing more, nothing less. If any of the commands
exists, it will be used.

From the top of my mind, those commands are:

	/usr/bin/net			sharesmb
	/usr/sbin/exportfs		sharenfs
	/usr/sbin/ietadm		shareiscsi/IET
	/usr/sbin/tgtadm		shareiscsi/SCST

Some of the iSCSI modules creates proc entries:

	/proc/net/iet/volume		shareiscsi/IET
	"/proc/net/iet/session		shareiscsi/IET
	/sys/kernel/config/target	shareiscsi/LIO

But all that is automatic - if they exists, they will be used, if not
ignored..

So I suggest that the "verified to work" is a five minute test - install
the dist, install ZoL, create a pool, create a volume/filesystem, set the
share property and check that it worked.

>> Despite my concerns above I'm willing to revisit merging the existing
>> sharing patches

If I got a penny for each time you've said that, I might not be a millionaire
but I would at least have gotten a free ice-cream from it! As I see it,
Brian isn't trustworthy enough.

>> as long as they meet the same requirements of as other proposed changes.

They do and they have all along.

> Very good.  Then we have an opening to try to reduce the difference, and
> that is good.  My first priority would be the NFS patch, as I use NFS
> (and not SMB and iSCSI) myself.  I suspect the SMB patches are important
> to more people, thought.

Just don't hold your breath.

>> * They should be rebased against the master branch, apply cleanly, and
>> introduce no new build warnings or failures as reported by the
>> automated testing.
> 
> When you say 'automated testing', do you mean something done with each
> pull request, or something that need to be run by the requester outside
> github?

ZoL have a automated build and test system which run on each commit.

These run on a plethora of distributions and versions. Have a look at

  http://build.zfsonlinux.org

>> * They should update and enable all of the existing 'share' tests in
>> the ZFS Test Suite.  These tests will be run by the automated testing
>> infrastructure and must to pass on all the regularly tested
>> distributions.
> 
> Right.  Where can we learn how to do this?  I know nothing about ZFS
> internals myself, but know how to program.

Look at the 'tests' directory. There's one test case per file in
'tests/zfs-tests/tests'. Just grep for 'sharenfs' and 'sharesmb'.

The original, which the ZoL dir was ported from, is

  https://github.com/illumos/illumos-gate/tree/master/usr/src/test/zfs-tests

Comparing the two, I can see that Illumos don't have any tests for
iSCSI either. So we need to design those. But see bottom of mail..

>> * They must receive positive code reviews from a ZFS developers other
>> than the author and myself.
> 
> This sound like an impossible requirement to me, given that no-one among
> the ZFS reviewers cared enough about the relevant patches for several
> years to bring the patches forward?  What make it likely that anyone
> else will look at it?

Indeed! And some of them have even indicated a serious aversion about
having sharing in ZFS/ZoL in the first place (as Carlos), so I'm not
expecting that to be happening :(



Now, because of Brians untrustworthiness in keeping his promises, I'm
very unveiling to jump through hoops and then realize I've wasted even
more time on this!! Keeping the patches up-to-date is easy, there's
no work to be done there (because they don't really touch anything that's
modified, so there's no/very little conflicts).

So I say, the code is there, it can be reviewed WITHOUT any extra work
from anyone. If he can do that, and give an actual review, that shows
enough willingness to do the job for me, and I can therefor continue
MY part.