Actually this was a question of priority. If Subversion assumes that "send_copyfrom_args" is stronger than "start_empty" (this simple idea just didn't come to my mind), then it's not a bug.
Originally I thought that if I report set_path "directory" revision=50 start_empty=TRUE I tell Subversion that I have no files within that directory at the revision I want to update to (but also I have no files at earlier revisions within that directory). Now with your comment I see that this is not true about earlier revisions. What about code stability: from my experience "send_copyfrom_args" argument in update works perfectly, I don't expect that it has many bugs. Thanks for the explanation! -- Dmitry Pavlenko, TMate Software, http://subgit.com/ - git+svn on the server side В сообщении от 3 August 2012 01:49:47 автор Philip Martin написал: > Dmitry Pavlenko <pavle...@tmatesoft.com> writes: > > As I understand, by telling start_empty=TRUE I say to the remote part > > (server) that I have no files in path "", and revision, specified to > > "set_path" is ignored. So I expect that Subversion won't send me > > "copyfrom_path" in "add_file" editor call. > > > > But it does, and this is a bug to my opinion. > > send_copyfrom_args was added to Subversion a long time ago but has never > been used by the rest of Subversion because the copyfrom information is > not useful in most cases. Since it is rarely used and not regression > tested it may have bugs. That said, the behaviour you report doesn't > look like a bug. If you ask for copyfrom information to be sent why > would it be a bug to send it? It may not very useful information but if > you ask for it why do you expect Subversion not to send it?
