I have the following problem using subversion:
I'm currently working on the trunk of my project and plan to do some refactoring (which includes renaming files or moving files to different directories).
At the same time someone else is working on the same project on a branch.
At some time I want to merge the changes made on the branch back to the trunk. That includes changes made to files (on the branch) that have been renamed on the trunk.
I did some tests and it seems that either subversion is not capable of following these changes or I'm missing someting (which is what I hope for). I tested this using the following script (should work in bash, assumes an svn repository at "http://myserver/svn/sandbox"):
svn co http://myserver/svn/sandbox cd sandbox/ mkdir -p MyProject/trunk MyProject/branches MyProject/tags cat - <<EOF >MyProject/trunk/FileOne.txt Test 1 2 EOF svn add MyProject svn commit -m "init" # create a branch svn copy http://myserver/svn/sandbox/MyProject/trunk http://myserver/svn/sandbox/MyProject/branches/Branch_1 svn copy http://myserver/svn/sandbox/MyProject/trunk http://myserver/svn/sandbox/MyProject/branches/Branch_1 # rename the file svn move MyProject/trunk/FileOne.txt MyProject/trunk/FileTwo.txt svn commit -m "renamed file" svn update # change the content of FileOne in branch cat - <<EOF >MyProject/branches/Branch_1/FileOne.txt Test 2 3 EOF svn commit -m "changed branch" # I now try to merge the changes in FileOne back to FileTwo cd MyProject/trunk/ svn merge -r1:HEAD http://myserver/svn/sandbox/MyProject/branches/Branch_1 # but this yields the following message: # Skipped missing target: 'FileOne.txt'
Any help is greatly appreciated.
Edit: Perhaps the process suggested by mikegrb could by somewhat automated by first generating a map of renamed files (old->new) from the svn log command on the trunk:
svn log -v ------------------------------------------------------------------------ r33 | sme | 2008-10-09 15:17:54 +0200 (Do, 09 Okt 2008) | 1 line Changed paths: D /MyProject/trunk/FileOne.txt A /MyProject/trunk/FileTwo.txt (from /MyProject/trunk/FileOne.txt:31) resulting map: {FileOne.txt => FileTwo.txt}
Now use this map to change filenames in the patch file generated on the branch.
Original:
Index: FileOne.txt =================================================================== --- FileOne.txt (.../trunk) (revision 31) +++ FileOne.txt (.../branches/Branch_1) (revision 34) @@ -1,3 +1,3 @@ Test -1 2 +3
modified:
Index: FileTwo.txt =================================================================== --- FileTwo.txt (.../trunk) (revision 31) +++ FileTwo.txt (.../branches/Branch_1) (revision 34) @@ -1,3 +1,3 @@ Test -1 2 +3
Just an Idea, haven't done it yet.
You want to checkout a working copy of trunk and then use the svn merge --reintegrate option: $ pwd /home/user/project-trunk $ svn update # (make sure the working copy is up to date) At revision <N>. $ svn merge --reintegrate ^/project/branches/branch_1 --- Merging differences between repository URLs into '.
This basic syntax— svn merge URL —tells Subversion to merge all changes which have not been previously merged from the URL to the current working directory (which is typically the root of your working copy).
I think this is an existing subversion bug - but don't hold your breath, its been open since 2002.
Unfortunately, this is one of the limitations of subversion. When we recently had a similar situation to deal with our solution was to create one giant diff for the branch and then go through it file by file manually patching trunk. Files that have been renamed and are not found will cause patch to prompt for the file name to patch. Very sub optimal. Watch out for binary files which won't show in the diff. This was one of the big factors that prompted us to evaluate other version control systems and ultimately decide to transition to git.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With