Speed on Windows for big repos

Matt Mackall mpm at selenic.com
Mon Jan 28 11:55:33 CST 2008


On Mon, Jan 28, 2008 at 10:37:28AM +0100, Adrian Buehlmann wrote:
> Some possibly very stupid/philosophical questions from a newbie like me 
> might be:
> 
> Why are flush() function calls needed at all (also in the UNIX case)?

The usual case is that we write some data via one file handle (held
open) and read it via another. Because the buffers associated with the
file are internal to the application and it's not aware that the file
handles point to the same file, we need to flush out the write side to
the OS level.

> Shouldn't that be an internal detail of a lower level abstraction?

Arguably, we should just kill the application-side buffering in these
cases.

-- 
Mathematics is the supreme nostalgia of our time.


More information about the Mercurial-devel mailing list