On Wed, Aug 29, 2007 at 11:19:58AM +0100, Simon Marlow wrote: > What I mean by "worse things could happen", is that darcs doesn't get a > chance to clean up when you hit Ctrl-C. I don't know a great deal about > darcs internals, but I guess it relies on being able to clean up in order > to leave the repo in a consistent state. Maybe it's always possible to > recover from a catastrophic failure using 'darcs repair', I don't know.
On Windows, darcs repair isn't always sufficient, since there's no atomic file rename over an existing file, so there's always a small window between when we remove the old inventory and when we rename the new. But that *should* be a very small window indeed, and usually repair should restore consistency. I have some ideas that should massively reduce the sensitive time during which we wish to block signals (and interrupts), but that'll come after the conflicts stuff is coded up, unless someone beats me to it (and there was an interested developer on the project). -- David Roundy http://www.darcs.net _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
