More digging turned up ...
On 7/1/06, Erik Huelsmann <ehuels@gmail.com> wrote:
> For performance and correctness reasons, I think we should stop
> running merge twice in update/switch.
>
> In the past, arguments were made that we don't run the actual
> (possibly interactive) diff3 in the middle of a network connection to
> be sure that the network doesn't time out. After I started
> investigating this issue last week, I found
>
> * That we actually run the external (possibly interactive) diff3 in
> the close_directory phase of the editor drive, which is as much 'in
> the middle of the connection' as close_file is (from which merge_file
> is called).
... that jpieper added the dry-run merge in r9868 to be able to do
per-file progress reporting when he switched libsvn_wc from per-file
to per-directory log running (to make libsvn_wc more efficient at
entries rewriting).
My change enables per-file progress reporting while still doing
per-directory log-running without a dry-run. So, I guess it's OK to
commit, apart from any other comments...
bye,
Erik.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Jul 1 11:11:04 2006