[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Showing apt progress info in apt-setup?



On Saturday 05 January 2008, Frans Pop wrote:
> On Saturday 05 January 2008, Joey Hess wrote:
> > > Also, if I select <Cancel> (press <enter>) while a Packages file is
> > > being downloaded (from the slow mirror), nothing happens: the
> > > download just continues, the mirror is added and apt-setup continues
> > > to the "services" dialog (using priority medium).
> >
> > IIRC this worked in my testing.
>
> Well, it doesn't now. Could you test again?

I really don't see how it can ever have worked as you test for an error 
value of 30, while debconf-apt-progress is being run inside in-target and 
that has:
if [ "$ERRCODE" != 0 ]; then
        exit 1
else
        exit 0
fi

That does not change the fact though that AFAICT debconf-apt-progress does 
_not_ exit 30, but just 0 when cancelled.

Hmm. Just trying again with the Indonesian mirror using a BC image, and it 
now refuses to cancel again. But I seem to see it cancel on other occasions 
(for example when I rerun apt-setup a second time).
/me is confused but strengthened in his conviction that cancellation support 
needs more testing and almost certainly additional changes.

I also wonder if we really need to have the cancel option enabled for _all_ 
generators. For example, it seems unnecessary and could possibly cause 
errors while scanning CDs.
Maybe we should only enable/disable the cancel capability _inside_ the 
generators and only for the specific commands where we want it.

Finally a minor usability issue. Showing the progress for downloads means 
the progress bar texts that show what generator is being executed are only 
displayed for a very short time. This means that 


Given these open issues I'm going to do an upload based on the current code 
minus the progress/cancellation changes, but will commit them again 
afterwards.


Finally, I have just committed some additional changes that make 
apt-setup-verify interpret the --from/--to options itself instead of just 
passing it on. This seems a simpler implementation and allows to give each 
source line to be tested an equal part of the passed range.

Cheers,
FJP

Attachment: signature.asc
Description: This is a digitally signed message part.


Reply to: