Hi there, I'm currently preparing our move from CVS to SVN. We've got a rather large application consisting of several modules some of which are specific for some customers.
Our repository will look like this: /trunk/core-module /trunk/xyz-module /trunk/customerABC-module-1 (several modules per customer) /trunk/customerDEF-module-1 /branches/customerABC/solution1/1.0/core-module /branches/customerABC/solution1/1.0/xyz-module /branches/customerABC/solution1/1.0/customer1-module1 /branches/customerABC/solution1/1.1/core-module /branches/customerABC/solution1/1.1/xyz-module /branches/customerABC/solution1/1.1/customer1-module1 /branches/customerDEF/solution2/1.0/core-module /branches/customerDEF/solution2/1.0/xyz-module /branches/customerDEF/solution2/1.0/customer2-module1 So, we only get some of the modules (read: subdirectories) of /trunk in the customer branches. When converting using cvs2svn it generates SVN commits like this: add /branches/customerABC/solution1/1.0/ (copy from /trunk) delete /branches/customerABC/solution1/1.0/customer2-module Of course, when reintegrating to trunk, I'm guaranteed to get a tree conflict if anything has been changed there (and if not, the module gets deleted which is even worse). After the move to SVN, merging will not be a problem since I've created a wrapper script which allows us to branch only certain modules (using SVN copy with multiple sources). This has the disadvantage that the topmost branch directory like /branches/customerDEF/solution2/1.0/ is not a parent of /trunk/ but we can work around that by applying "svn merge" to each module sequentially, then committing everything at once. So back to my original question: How do I manage these kinds of merges? - How do I avoid tree conflicts when merging trunk to branch and there are changes in a module not present in the branch? - How do I avoid merging back the "delete module" operation generated by cvs2svn? Thanks, Tino. -- "What we nourish flourishes." - "Was wir nähren erblüht." www.tisc.de
