darcs

Issue 2130 rollback keeps prompting when it should abort

Title rollback keeps prompting when it should abort
Priority Status given-up
Milestone Resolved in
Superseder Nosy List simon
Assigned To
Topics

Created on 2012-01-16.14:58:15 by simon, last changed 2017-07-31.00:33:59 by gh.

Messages
msg15025 (view) Author: simon Date: 2012-01-16.14:58:14
<sm> I always find rollback confusing. What's happening here ?                                                                                                                                                                              
     https://gist.github.com/7497621babb8cc5b5f35                       
<sm> I guessed it found 62 dependencies, but then I'd expect the rollback to                                                                                                                                                                
     abort if I answer N to any of them ?
<Heffalump> sm: I agree, but I may also be confused                     
<Heffalump> sm: oh, I see.                                              
<Heffalump> some of the dependencies are independent of each other
<Heffalump> if you answer N to any of them, it'll never ask you about the last                                                                                                                                                              
            patch which is the one you were actually targeting          
<Heffalump> so if you have X depending on Z and Y depending on Z, asking to                                                                                                                                                                 
            rollback Z will ask you about X,Y,Z. If you say n to X, then it'll                                                                                                                                                              
            still ask about Y unless X also depends on Y.
msg15027 (view) Author: ganesh Date: 2012-01-16.17:09:47
This is a general feature of patch selection, for what it's worth - 
e.g. obliterate will exhibit exactly the same behaviour, as will pull 
albeit in the other order.

What do you think should happen instead? My feeling is that the 
behaviour needs to be consistent when one patch is selected or when 
multiple patches are selected, which I think complicates the obvious 
fix of stopping when there's any 'n' response. I guess we could stop 
when all of the explicitly selected patches are ruled out.
msg15030 (view) Author: simon Date: 2012-01-16.17:34:49
Operating on a single target patch, I'd expect the command to recognise as 
soon as the operation on that patch has become impossible and stop at that 
point. Which multiple target patches - I'm not sure, what's an example of 
that case ?
msg15031 (view) Author: simon Date: 2012-01-16.17:35:57
*With* multiple target patches...
msg15033 (view) Author: ganesh Date: 2012-01-16.18:39:02
On 16/01/2012 17:35, Simon Michael wrote:
> 
> Simon Michael <simon@joyful.com> added the comment:
> 
> *With* multiple target patches...

In your example from IRC, suppose darcs changes -p dumb had reported
multiple patches.

Ganesh
History
Date User Action Args
2012-01-16 14:58:15simoncreate
2012-01-16 17:09:48ganeshsetmessages: + msg15027
2012-01-16 17:34:50simonsetmessages: + msg15030
2012-01-16 17:35:57simonsetmessages: + msg15031
2012-01-16 18:39:02ganeshsetmessages: + msg15033
2017-07-31 00:33:59ghsetstatus: unknown -> given-up