Branko Čibej wrote on Fri, 05 Jul 2013 22:12:34 GMT: > Can you confirm that all the branches created by copying (svn copy) from the > source, as you've described in this diagram? Damn you are right. Some unexperimented people thought years ago that they were more clever than the system, created empty folders at the branches root, and then copied some sub-folders into them...
So svn 1.8 reluctancy to do the merges seems legitimate in these cases. I'm currently investigating the details and implications of these horrors, but I feel we are a bit doomed. I fear we will have to rework the layout of our repository, in order to reset properly the branches ancestry. Daniel Shahaf wrote on Sat, Jul 06, 2013 at 18:00: > I believe '--ignore-ancestry' is supposed to override this particular check. Well, I already tested that, with and without --force, without more success. > The code has changed on 1.8.x@HEAD since 1.8.0[1], so 1.8.1 will behave > differently. (It will probably just issue a different error message, since > [1] claims that it didn't change the checks being done.) So I just compiled a svn (on trunk@1500602), and get the following errors: $ svn --version .... svn, version 1.9.0-dev (under development) .... $ cd svn/v3-dev $ svn merge ^/app/v2-dev svn: E195016: 'svn://......./app/v2-dev@70081' must be ancestrally related to 'svn://......./app/v3-dev@70081' $ svn merge ^/app/v2-dev --ignore-ancestry svn: E200004: Cannot merge automatically while ignoring mergeinfo We'll continue our investigations, and keep you updated. Thanks -- Eric Estievenart
