2015-06-02 9:35 GMT+02:00 Andrew Church <achu...@achurch.org>: > You would be wrong, because patch already creates exactly one additional > copy of the file (*.orig). Formerly, foo.orig was always the original > directory entry and foo was always a newly created file. With the > behavior I argue to be a bug, if all hunks for a file fail, then foo is > the original directory entry and foo.orig is a newly created file. > Either way, patch uses the same amount of disk space.
That's indeed weird; the .orig file would have to be a hardlink for the kind of consistency I'm thinking of. But then that might confuse people who are not used to hard links. (It's a different situation if you're working with hard- linked trees in the firt place.) I'll think about this some more. Thanks, Andreas