Paul Burba wrote:
> On Thu, Jun 5, 2008 at 2:29 PM, Paul Burba <ptburba_at_gmail.com> wrote:
>> On Fri, May 9, 2008 at 12:02 PM, Kamesh Jayachandran <kamesh_at_collab.net> wrote:
>>> Hi All,
>>> I hereby attach the fix to issue 3067.
>>> I am bit skeptical about performance of 'guess_live_ranges' in this patch.
>>> But it would perform poorly only for the situations when
>>> merge_src_subtree_at_revision2 does not exist and that too for one invocation.
>>> Would like to know what others would think about it.
>>> With regards
>>> Kamesh Jayachandran
>> Hi Kamesh,
>> Sorry for the long delay, while working on this I've found other
>> related problems (issue #3199 and #3157) that needed to be addressed
>> first. But *finally* I reworked our patch, using a simpler approach
>> that avoids the round trip overhead of guess_live_ranges() entirely.
>> But first I should mention a few things:
>> 1) I added/expanded a couple of merge tests in r31575 to cover a wider
>> range of scenarios where the subtree of a target needs different
>> ranges applied than the target. Our existing tests were lacking in
>> this area.
>> 2) In r31594 and r31601 I made some improvements that solve some of
>> the problems in 1). The core problem of 3067, where we describe
>> describe invalid subtrees to the merge report editor, still remains
>> and this attached patch addresses that.
>> Ok then, there are two key parts to this patch:
>> The first is the new helper for filter_merged_revisions() called
>> prepare_subtree_ranges(). See the doc string for
>> prepare_subtree_ranges for an explanation of what I'm trying to do.
>> Please give some thought to the question posed in the section titled
>> '### This is tricky, sort of the inverse of B' as that is the one
>> thing about this patch I'm not entirely confident in.
>> The second is in filter_merged_revisions(), see the code in the block
>> 'A little trick: If CHILD is a subtree which will be deleted by'.
>> Let me know what you (or anyone else interested in looking at this) think.
> Hi Kamesh,
> As discussed today in IRC, here is my latest version of the patch
> which now addresses the problems described in
With your latest patch I was able to do the merge from trunk into the
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-06-26 20:24:41 CEST