Re: paths in diff output (possible bug)

2012-05-15 Thread Stefan Sperling
On Tue, May 15, 2012 at 10:48:13AM +0200, Stefan Sperling wrote: > Ah, yes, you are right. We should filter deleted props, too. Done in r1338688.

Re: paths in diff output (possible bug)

2012-05-15 Thread Stefan Sperling
On Tue, May 15, 2012 at 01:16:44AM +0200, Dmitry Pavlenko wrote: > Hi Stefan, > > I'll have a look a bit later. But from the first glance I've found a couple > of suspicious places. I > do not state that they are wrong or that I have a counterexample, but just > not clear for me. > Thanks muc

Re: paths in diff output (possible bug)

2012-05-14 Thread Dmitry Pavlenko
Hi Stefan, I'll have a look a bit later. But from the first glance I've found a couple of suspicious places. I do not state that they are wrong or that I have a counterexample, but just not clear for me. 1. When you call callbacks->dir_props_changed for added properties, you convert them to r

Re: paths in diff output (possible bug)

2012-05-14 Thread Stefan Sperling
On Thu, May 03, 2012 at 10:05:26PM +0200, Dmitry Pavlenko wrote: > Thanks, Stefan, > I like the approach of a fake checkout. Actually this is what I meant by > proposing to handle this > case separatedly. > Hi again Dmitry, can you please check if r1338291 fixed diff output as expected? Thanks

Re: paths in diff output (possible bug)

2012-05-03 Thread Dmitry Pavlenko
Thanks, Stefan, I like the approach of a fake checkout. Actually this is what I meant by proposing to handle this case separatedly. > On Thu, May 03, 2012 at 02:51:54PM +0200, Stefan Sperling wrote: > > I don't want a solution that requires a server upgrade. > > This is a user-interface problem.

Re: paths in diff output (possible bug)

2012-05-03 Thread Stefan Sperling
On Thu, May 03, 2012 at 02:51:54PM +0200, Stefan Sperling wrote: > I don't want a solution that requires a server upgrade. > This is a user-interface problem. The client/server interaction > is capable of producing the required information. We just have to > display the information in a better way.

Re: paths in diff output (possible bug)

2012-05-03 Thread Stefan Sperling
On Thu, May 03, 2012 at 02:43:11PM +0200, Dmitry Pavlenko wrote: > Helllo Stefan. > > Thanks for trying to solve this problem. > Unfortunately you patch doesn't solve the problem because of 2 reasons. > 1. It doesn't change "target" (in diff request), and it still may contain '/'. > 2. If in my ex

Re: paths in diff output (possible bug)

2012-05-03 Thread Dmitry Pavlenko
Helllo Stefan. Thanks for trying to solve this problem. Unfortunately you patch doesn't solve the problem because of 2 reasons. 1. It doesn't change "target" (in diff request), and it still may contain '/'. 2. If in my example (just replace http://localhost/svn == file:///tmp/test) diff fails if

Re: paths in diff output (possible bug)

2012-05-02 Thread Stefan Sperling
On Tue, Mar 20, 2012 at 07:16:44PM +0100, Dmitry Pavlenko wrote: > > > I expected you'd get an "/directory/subdirectory@1 doesn't exist" error. > > I was the older (1.6) SVN behaviour > > Now "do {} while()" cycle in "diff_prepare_repos_repos" takes parent URL > until URL exists in both > star

Re: paths in diff output (possible bug)

2012-03-20 Thread Dmitry Pavlenko
> I expected you'd get an "/directory/subdirectory@1 doesn't exist" error. I was the older (1.6) SVN behaviour Now "do {} while()" cycle in "diff_prepare_repos_repos" takes parent URL until URL exists in both start and end revisions. So maybe the cycle should be run only once not to let '/'.

Re: paths in diff output (possible bug)

2012-03-20 Thread Daniel Shahaf
Dmitry Pavlenko wrote on Mon, Mar 19, 2012 at 21:54:58 +0100: > Hello all. > I'm trying to understand the principles how SVN calculates paths for diff. > > I've put repository attached under URL "http://localhost/svn";. I'll put its > history here: > --