Hi Osamu,
On Wed, Mar 09, 2016 at 11:24:09PM +0900, Osamu Aoki wrote:
> On Wed, Mar 09, 2016 at 01:38:46PM +0100, Andreas Tille wrote:
> ...
> > > It changed when a backport of devscripts >= 2.15.10 was installed on the
> > > QA systems.
> >
> > Thanks for the very helpful explanation. So if I would like to restore
> > the old behavior I would need to discuss the strings in a bug report
> > against uscan, right? From my naive point of view the changes of the
> > strings is not a good idea (besides the side effect in UDD). What do
> > others think?
>
> In retrospective for backport, string should have been kept if possible.
As Raphael said changing is not a good idea for several reasons. My
additional point is that the strings become worse to understand. Could
you give good reasons for the change? If not, please revert it.
> In old code, error reporting had several irregularities which got
> cleaned up in new uscan. This reorganization was quite extensive
> so I should have warned James when he started working on it.
I admit I'm quite happy about the new uscan - just those strings are
misunderstanding in themselves, sorry.
> Now uscan reports more detailed error analysis, please allow this new
> feature update which already happened.
The feature is fine - please *exactly* describe why the string
'newer package available'
should be better than
'Newer version available'
IMHO this is contextual wrong since there is no new package yet when
uscan is parsing a new version.
> > PS: I have changed the web sentinel to accept both strings, so from a
> > Blends point of view this is should not be an issue any more.
>
> Thanks.
That's needed in any case since the time for discussing this and the
time a potentialy changed uscan might be installed on udd.d.o would last
to long with a broken output.
Kind regards
Andreas.
--
http://fam-tille.de
