os/bluestore: pass strict flag to bluestore_blob_use_tracker_t::equal()

[xiexingguo]

And if that flag is true, we'll do a strict equivalence comparison instead.

Signed-off-by: xie xingguo xie.xingguo@zte.com.cn

[ifed01]

I'm disagree with the initial assumption that <1> isn't equal to <2>. In fact <2> doesn't have that actual range information [1..1001] - it contains total referenced bytes counter only. And that amount of referenced bytes equivalent (weakly?) to the one at <1> hence returning true. I.e. equal method returns true if there is a chance that two trackers describe the same used_range rather than strict comparison. Probably it makes sense to change the function name though..

[xiexingguo]

I.e. equal method returns true if there is a chance that two trackers describe the same used_range rather than strict comparison.

I see the only caller is fsck(), which means we use this method to measure the same blob, so I guess we'd better narrow the constraints down.

[ifed01]

I'd prefer to leave current comparison mode as well. May be we can have a boolean flag for equal() call to request strict/weak equivalence comparison.

[xiexingguo]

I'd prefer to leave current comparison mode as well. May be we can have a boolean flag for equal() call to request strict/weak equivalence comparison.

Updated!

[ifed01]

Could you please add some UT for that

[ifed01]

LGTM.

[xiexingguo]

Could you please add some UT for that

Sure, added.

[xiexingguo]

http://pulpito.ceph.com/xxg-2017-09-01_07:17:41-rados-wip-xxg-testing-2017-09-01-1357-distro-basic-smithi/
http://pulpito.ceph.com/xxg-2017-09-03_12:51:11-rados-wip-xxg-testing-2017-09-01-1357-distro-basic-smithi/

[xiexingguo]

retest this please

[xiexingguo]

retest this please