On 4/27/06, Peter N. Lundblad <peter@famlundblad.se> wrote:
> Ivan Zhakov writes:
> > On 4/26/06, Peter N. Lundblad <peter@famlundblad.se> wrote:
> > > Ivan Zhakov writes:
> > > My reasoning is that it is so harmless if it happens that actually reporting
> > > it makes more harm by confusing users, but I'm not really objecting.
> > >
> > But if sigfault happens after wcprops directory have deleted we loose
> > wcprops at all.
> > Of course I like simpler code, so if we don't fear such situations
> > I'll do it as you suggesting.
> >
> Oh, I was expressing myself confusingly today. I think we should remove
> the props between writing the log and running it (as you suggested), so we
> don't risk loosing them. The problem is that we risk leaving the old props
> behind if an error happens here. I think we can live with that, and if we do
> we could as well ignore errors when trying to delete the wcprops to not confuse
> users in this case. The result after cleanup will be the same.
Ok, I'll do it so tomorrow.
>
> A reason I want to keep this upgrade simple is that it is hard to
> test (at least with our test suite), so it is good if it is simple.
Agreed with you.
--
Ivan Zhakov
Received on Wed Apr 26 22:23:30 2006