Re: [Quilt-dev] quilt diff and stderr

[Peter Williams]

Jean Delvare wrote:
> Hi Andreas,
>
> [Jean Delvare]
>
> > > The "diff" command of quilt currently prints a message on stderr in
> > > certain circumstances:
> > >
> > > More recent patches modify files in patch X
> > >
> > > And this is true, but I don't quite see how this constitutes an
> > > error.
> > >
> > > This seems to be more of an additional information of arguable
> > > interest than an error. Or am I missing something?
>
>
> [Andreas Gruenbacher]
>
> > Yes, it's not an error but only a warning. No error is returned.
>
>
> Actually, an error value (1) *is* returned, so maybe we could fix at
> least that if you think it's not correct. What I am personally concerned
> about here is stderr though, not the return value.
>
>
> > Maybe we could make that more obvious. I disagree with your
> > assessment that the information is irrelevant though: the user may
> > have modified some of the  files that patches further on top of the
> > stack modify as well, expecting the  changes to show up in the below
> > patch, but they won't. This could cause a lot of confusion.
>
>
> I don't understand. Could you please detail a concrete scenario where
> this message would appear and would be desired?
>
>
> > The diff command is also about seeing what will be in a patch after
> > the next refresh.
>
>
> The manual page doesn't say that. Maybe we should make it clearer? I
> have always wondered whether "quilt diff" was getting the data from
> patches/* of from .pc/* and the current files. It would be the latter if
> I read you correctly.

While you're mucking around with return values I'd like to request that 
the return values be made more informative.  This will help me to handle 
them sensibly in gquilt without having to parse the stderr output (which 
will break when the output is localized to something other than English).

The kinds of thing I would like to be able to determine from the error 
return value are:

1. Is the use of -f (for force) being recommended?
2. Is a refresh and retry being recommended?
3. Both?
4. Neither?

This allows me to offer the user a button(s) to press to either carry 
out the recommended action on their behalf or cancel the operation.  I 
think that this greatly improves gquilt's usability.

Thanks
Peter
--
Peter Williams                                   address@hidden

"Learning, n. The kind of ignorance distinguishing the studious."
 -- Ambrose Bierce