On Jun 19, 2013, at 12:08 , Philip Martin wrote: > Philip Martin <philip.mar...@wandisco.com> writes: > >> Ben Reser <b...@reser.org> writes: >> >>> I don't imagine it'd take very long at all to implement but the >>> problem of course is that we really should think carefully how we go >>> about doing this. If we can detect this at runtime we probably >>> should. >> >> I don't see how Subversion can determine that one script needs a >> terminal while another can use a file, only the user knows that. >> >> It's also hard to fix 1.8, how do we pass the information into the >> client library without changing the API? Perhaps we could recognise a >> special part of the command name or a special external parameter, so >> >> --diff-cmd svn:interactive:myscript >> >> or >> >> --diff-cmd myscript -x svn:interactive >> >> gets a terminal while >> >> --diff-cmd myscript >> >> gets a file. > > Or we could extend the opaque svn_stream_t to make the underlying > apr_file_t available. Is mixing output to the file and output to > the stream acceptable or does it introduce output order problems?
I tried & installed both your patches and both work just fine, thank you very much. Without giving it much thought, I now went for your second patch (it seemed to be the more sophisticated one). I think something like the "-x svn:interactive" option would be the best solution, since that way the diff-cmd really just remains a command. Thank you very much for your quick help! I'll definitely keep tabs on the development of this, I hope that one way or another it will make it into the svn code base. Michael
smime.p7s
Description: S/MIME cryptographic signature
