Re: fsync for only part of a file

On Sun, 14 Sep 2003, wrstuden-DZEk9q7Sfju/3pe1ocb+swC/[email protected] wrote:
> On Sat, 13 Sep 2003, Alan Barrett wrote:
> > This looks more like a variant of fdatasync() than a variant of fsync().
> 
> I don't think so. I think fsync_range() is better than what I suggested,
> but I fully expect that the metadata, at least what's needed to read the
> sync'd range, would get flushed. i.e. my thought is against cutting the
> fdatasync() corner. :-)

I agree with all that.

I was thinking of atime and mtime changes, which I would expect to be
flushed by fsync() but not by fdatasync(), and which probably do not
need to be flushed by fsync_range() in the normal case.

--apb (Alan Barrett)

---------------------------------------------------------------------
The BSD APIs Discussion Mailing List
To unsubscribe:
send "unsubscribe bsd-api-discuss" to majordomo-of7zbby7T3pWk0Htik3J/[email protected]