< Previous by Date Date Index Next by Date >
< Previous in Thread Thread Index Next in Thread >

Re: [reSIProcate] merge mayhem?



On 27/09/13 23:16, Philip Kizer wrote:
> On 2013, Sep 27, at 14:52, Daniel Pocock <daniel@xxxxxxxxxxxxx> wrote:
>> Looks OK, but the commit mail shows something else, notice the
>> contrib/cajun/include stuff appears to be both added and removed again
>> in the same commit:
> 
> I know there have been _many_ merge-related SVN changes since then, I'm 
> wondering if that's possibly related to this:
> 
> http://svn.apache.org/repos/asf/subversion/trunk/CHANGES
> 
> Version 1.7.5
> (17 May 2012, from /branches/1.7.x)
> http://svn.apache.org/repos/asf/subversion/tags/1.7.5
> 
>     * fix merges which first add a subtree and then delete it (issue #4166)
> 
> 
> Regardless, 1.6 is quite old (2011/06/01), so I'd like to do some tests to 
> verify if it's client-side or server-side.
> 
> 
> I have the new dump+restore repo that I'm close to ready to swap out still 
> just waiting on when the repo is a bit more quiescent, but I can trivially 
> make hot-copies of it for your testing. [For everyone else, the SVN server 
> software for resiprocate has been updated to stay current, but the repo 
> format is still from an older version. I've dumped and restored it on an 
> alternate directory and I'll want to swap that newer version in at some 
> point, but I think that might require client checkout refreshes when I do so, 
> still investigating and testing to be sure before announcing).
> 
> It's current as of 10540 and I can easily make it current as of 10555, 
> immediately prior to the merge you're mailing about.
> 
> 
> Would you like to test with your current 1.6.x client to see if the behaviour 
> is the same with a newer backend database?  If so, I'll not wait on 
> quiescence on the repo to get that put in.  [Though I will still wait until I 
> know if you will need to do some explicit refresh since it does alter some of 
> the metadata because of the more efficient storage heuristics.]
> 
> If the behaviour is the same then I'd then like you to do another test with a 
> newer client (preferably an up-to-date 1.8.3 client) to see if that shows the 
> same weird add-and-delete in the same commit.
> 

I tried making a checkout of the test URL using a Fedora 19 system, it
has SVN 1.7.11, it gives me an error though:

$ svn checkout https://svn.resiprocate.org/rep/resip-test/main resip-test

svn: E000002: Unable to connect to a repository at URL
'https://svn.resiprocate.org/rep/resip-test/main'
svn: E000002: Could not open the requested SVN filesystem


I used the SVN client on Fedora to merge b-gsoc-2013 into main and that
appears to have worked successfully.  It detected 4 tree conflicts and I
was able to resolve them and merge successfully.  I don't know if the
1.6 client would have reported the same tree conflicts as I didn't test it