Hello
I am currently having the same issue with merging two workspaces and
I noticed that no one has responded to this issue. I can't see how
the algorithm presented in JSR-170 (8.2.10.1) or the newly released
JSR-283 could possibly implement the textual description (8.2.10).
Nowhere in the algorithm is the merge result of the nearest
versionable ancestor taken into consideration. Specifically the
version of doleave(n) that dofail(n, v') calls needs to be altered to
read:
doLeave(n)
for each versionable child node c of n domerge(c).
rather than what is currently has:
doLeave(n)
for each child node c of n domerge(c).
thanks in advance
tim
On Jul 5, 2007, at 11:24 AM, Bob Wieler wrote:
From: Bob Wieler <[EMAIL PROTECTED]>
Date: Jul 5, 2007 11:24 AM
Subject: Merging non-versionable nodes
To: [email protected]
Hi,
I am hoping that someone can help me with what seems to be a
problem with the merging of non-versioned nodes (I am currently
using Jackrabbit 1.3). I am using a workspace that contains
versionable nodes with a number of non-versionable referenceable
children. After performing a merge the non-versionable children are
merged even if a conflict exists in the parent versionable node.
Unless I am greatly misunderstanding the JCR specification (both
the current JSR-170 and upcoming JSR-283) there appears to be a
conflict in the section describing merges and the algorithm
provided (which Jackrabbit implements).
From JSR-170 section 8.2.10 Merge:
"In either case, (regardless of whether bestEffort is true or
false) for each non-versionable node (including both referenceable
and non-referenceable), if the merge result of its nearest
versionable ancestor is update, or if it has no versionable
ancestor, then it is updated to reflect the state of its
corresponding node. Otherwise, it is left unchanged."
From this I read that if a versionable node has a merge result
other than update (a failure) then any non-versionable children of
that versionable node should be left unchanged (assuming bestEffort
is true of course). However the algorithm presented in 8.2.10.1 of
the specification seems to go against the above statement.
On a dofail(n, v') if bestEffort is true then the jcr:mergeFailed
property is set, the UUID is added to the failedset, and then
doleave(n) is done. doleave states that for each child node c of n
domerge(c) which in turn will perform a doupdate(n, n') for non-
versionable children.
Am I missing something fundamental about the way merging of nodes
should be done in JCR? If the merge of a versioned node fails
shouldn't it just perform a merge of the versioned child nodes
rather than all of the child nodes?
Thank you for your help,
Bob Wieler
