Re: move instead of copy
Julian Squires <[email protected]> Tue, 15 Sep 2015 17:51:51 -0400
| Newsgroups | gmane.network.unison.devel |
|---|---|
| Message-ID | <CAEbLoB+_2cGDJ2s9dY=aYCTv++TLXkZbG6yNT2HVwimsOjbCvQ@mail.gmail.com> |
I am considering implementing this. At least in the case of Posix systems, couldn't one rename(2) and if it fails with EXDEV, then fall back to copy-and-delete? On Tue, Sep 15, 2015 at 12:51 PM, Trevor Jim <[email protected]> wrote: > As I recall, the reason for the copy is that sometimes move is not > atomic. > > For example, a move from one hard disk to another is actually a copy and > a delete. The copy could fail in the middle. > > It is hard to tell how a move will be implemented from just the path of > source and destination. > > Using copy+delete instead of a move is therefore a safer approach, since > Unison knows that copies can fail and takes this into account so that > you do not lose data. > > Something to keep in mind... > > -Trevor > _______________________________________________ > Unison-hackers mailing list > [email protected] > http://lists.seas.upenn.edu/mailman/listinfo/unison-hackers -- Julian Squires