On 23/03/2008, Stefan.Fuhrmann_at_etas.com <Stefan.Fuhrmann_at_etas.com> wrote:
> "Simon Large" <simon.tortoisesvn_at_googlemail.com> wrote on 22.03.2008
> 23:49:32:
>
>
> > The description in changelog.txt shows some difference between the
> > handling of F5 in the rev graph and the log dialog.
> >
> > In log dialog I believe F5 just fetches the new HEAD and fills in any
> > missing log messages if the cache is not at HEAD. Both shift-F5 and
> > Ctrl-F5 will invalidate the cache (displayed messages only, all
> > messages for this path, all messages for this repository?) and refetch
> > the logs from the repository.
>
>
> Yes F5 will just check for a new HEAD revision and fetch the new
> revisions if they are relevant. It will also switch to on-line mode.
>
> Using <SHIFT> or <CTRL> (or both) in combination with F5 or the 'Refresh'
> button will switch the cache to on-line mode as well. But the log is
> fetched into a new, temporary cache object (log view content only - not
> the whole repository root history). Only if that was successful, the
> actual cache gets updated by adding new revisions and replacing the old
> revision info where there is new data in the temporary cache object.
> All other revision info remains unaltered because a full refresh of
> >100k revisions would take ages.
>
> Unfortunately, I just discovered two bugs that I will fix asap:
>
> * the accelerators don't specify the <SHIFT>, <CRTL> for F5
> * forced refresh should fall back to cached data when there is
> a network problem
>
>
> > In rev graph F5 does what shift/Ctrl-F5 does in the log dialog. Is
> > there any reason for them to behave differently? Different UI
> > behaviour makes it harder for people to remember what they should do.
>
>
> No, F5 in the revision graph will do the same as F5 in the log view.
> The only difference is that the revision graph requires the full history
> (i.e. calls 'svn log -v' on the repository root).
>
> But you are right, that there is an asymmetry: <SHIFT> and <CTRL> are
> ignored in combination with F5 because the *whole* history had to be
> fetched. IMHO, just having authors and log commens updated does not
> justify the risk here. The user may still call the log at a particular
> node and update the cache selectively from there.
OK, understood. I will update the docs to describe the behaviour as it is now.
Simon
--
___
oo // \\ "De Chelonian Mobile"
(_,\/ \_/ \ TortoiseSVN
\ \_/_\_/> The coolest Interface to (Sub)Version Control
/_/ \_\ http://tortoisesvn.net
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_tortoisesvn.tigris.org
For additional commands, e-mail: dev-help_at_tortoisesvn.tigris.org
Received on 2008-03-24 10:00:55 CET