Singlethreading patch series

[Nicolas Sebrecht]

On Mon, Jan 10, 2011 at 12:08:34PM +0100, Johannes Stezenbach wrote:
> 
> On Mon, Jan 10, 2011 at 11:00:17AM +0100, Sebastian Spaeth wrote:
> > 
> > I split the patches to be smaller and more self-contained and improved
> > the commit messages. While I had to touch the one "except:" part in
> > LocalStatus.py that we previously talked about, I decided to do
> > completely away with the Exception catching. Even catching
> > EnvironmentErrors as Johannes had suggested seems to much. If there is
> > an IOError or OSError when os.fsyncing() our status data base something
> > is really wrong and fishy and we should loudly complain rather than
> > silently ignore it.
> 
> While it is true that catching EnvironmentError would also
> catch IOErrors from fsync(), it would both fix your issue
> with KeyboardInterrupt *and* improve the current situation
> (no tracebacks and not aborting even for software bugs).
> Maybe it would be better (and suffcient wrt the intentions of
> the patch which introduced it) to only catch socket.error?
> 
> IMHO your patches 1...3 do not fix the no-traceback issue
> (leaving the catch-all except: + self.ui.warn in place),
> and litter the code with KeyboardInterrupt special handling.
> I would argue that catch-all except: is almost always a bug,
> and if you fix that we'd end up with a much better patch.

More precisely, this is the patch 2/4 which seems to defeat the purpose.

-- 
Nicolas Sebrecht