Hi Congwu,
On Aug 14, 2009, at 3:27 , Chen, Congwu wrote:
If I understand correctly, the existing code has a bug in implGetItem:
The assumption: "records detected new/changed now will inevitably be
new or changed in all other profiles" seems to me not reasonable.
Please
correct me if I lost something.
The assumption is correct because a sync remains resumable only if no
other profile is synced in between (should be that way, see update
below). This is because the implementation cannot hold resume states
for more than one profile (e.g. the chgl_markedforresume flag would
need to be per profile). This is of course a compromise, but I
considered that acceptable.
Because of this, your sample scenario cannot happen as described
Suppose we have record K and server A,B
1. sync with A, suspended.
2. K gets updated.
3. sync with B, K.syncmod=3, sending K.
At this time, syncing B will cause suspend status for A to be discarded.
4. sync with A, resume, K.syncmod=5
A will not be able to resume, but do a normal sync, and thus will not
update K.syncmod.
5. sync with B, K will be resended.
Because it was not updated in 4, K will not be resent.
Update: I found you are still right, because there seems to be an
implementation problem - the resume status of other profiles is not
cleared when a session completes ok. The idea was probably that the
resume state gets lost only when another profile's sync does not
successfully end, but after a quick review I guess this cannot work
properly. I also found that there might be a problem with pendingmaps
and multiple profile syncs, because these should be kept on a per
profile basis because these are needed not only in resumes, but normal
syncs as well.
I guess I'll have to do a more thorough review and cleanup of that
code under the aspect of multi-profiles.
Thanks a lot for your close observation and revealing the problem!
For single profile use, the code is still safe, IMHO.
Lukas Zeller ([email protected])
-
Synthesis AG, SyncML Solutions & Sustainable Software Concepts
[email protected], http://www.synthesis.ch
_______________________________________________
os-libsynthesis mailing list
[email protected]
http://lists.synthesis.ch/mailman/listinfo/os-libsynthesis
