On 10/6/13 5:29 PM, Gabriela Gibson wrote: > I can see Eitan's point -- if we went ahead and changed all the obsolete > <tt> tags on the Subversion website, it would remove a lot of useful > blame info.
No it doesn't. All the information is still there. It's harder to step back through revisions with the cmdline client because the display format is not suited very well for that. But you can still do it, just pass a revision one before the revision the line you're looking at changed in. A GUI could just as easily support a feature where you could walk through changes interactively to find where what you were looking for really happened (we could even make that much easier for the GUI to support than it is now). > Is a two stage operation possible? > > say, you type: > > 'svn commit --blame-revert ...' > > Part one would be a regular commit, part two would be a second commit that > restores all the original authors that the first commit modified; so one > commit would produce two consecutive revisions. > > To be (relatively) safe, this could be a repository admin action only. I honestly can't really fathom how you'd implement 'blame-revert'. Either a commit revision modified a line in the file or it didn't. Personally I'd much rather walk through changes and know that I've found what I wanted rather than hope that someone didn't tag something as "unimportant" that really was important for what I'm looking for. The idea that you can decide what a user far in the future is going to find important at commit time when running blame seems like a rather strange idea to me.
