[Cosmo-dev] Re: [Chandler-dev] Modification without dtstart report

[Jeffrey Harris]

Hi Jared,

> 4. Chandler should send a dtstart in EIM for modifications
> https://bugzilla.osafoundation.org/show_bug.cgi?id=9856

I think this is the crux of the problem, but it's certainly a good idea
for us to fix the other pieces of the puzzle that came crashing down
when this went wrong.

> If anyone has theories as to why we have not seen this "creating a
> modification apparently horks well-nigh everything" previously, I'd be
> curious to know.

After working through this with Randy, I'm 99% sure I introduced this
problem in r14914 last Thursday.  To make conflict resolution and change
merging work when a modification is removed from the server, I treat the
inbound deletion as if the remote state is actually all Inherit values
for EIM fields.  The resulting diffs work for Chandler's in memory round
trip tests because Chandler defaults absent attributes to Inherit.
Cosmo defaults them to None and we have a problem.

I'm working on a fix now.

Sincerely,
Jeffrey