Re: Bug#235732: Announce of the upcoming NMU for the atlas3 package
Greetings, and thank you again for your work and concern here!
Christian Perrier <bubulle@debian.org> writes:
> Quoting Camm Maguire (camm@enhanced.com):
> > Greetings, and thanks so much for your work here!
> >
> > As I understand your proposal, I do object, so please abort, at least
> > temporarily, so we can consult upon the best course forward. I have
> > already consulted the opinion of many atlas users, and the 'annoying
> > note opinion' is far from universally held. Please also note the
> > currently reduced priority of the notes. This said, I might be
> > persuaded that some or even all of the content belongs in
> > README.Debian. What I object to is basing this assessment on the
> > opinion of those submitting bugs, and not the vast silent majority who
> > appear to like things the way they are.
>
>
> Well, the announced NMU mixed indeed two topics, both of which being
> currently the subject of active actions by the i18n task force:
>
> -the multiplication of notes displayed by debconf
>
> -the proposed translations for *all* these notes, which are slowly
> being forgotten in the BTS
>
>
> Both are of course mixed: if notes are finally removed, entirely or
> partly, the translations are partly obsoleted.
>
> Both have something in common: the complete lack of visible maintainer
> activity in the BTS.
>
My apologies -- my available time comes in bursts.
> About debconf notes, all arguments that you raised could have been
> sent to the bug reports such as #235732...or the recent ones which I
> automatically raised during the mass bug filing about "useless" notes.
>
I suppose I should have done that -- my point was that contented users
are not going to.
> Some of the translations are sitting in the BTS for about 1.5 years
> without any comment.
>
> *this* is the initial point of this action: get some reaction about
> this. The minimal one would be an upload with all these translations
> fixed.
Certainly. I have a local upload waiting a fix for the custom build
target. make has broken/changed since atlas3 was put together, and
now the ordering of the rules is out of control. My intention was to
upload a fix for everything at once since it takes so long to get
atlas through the build system. This is the most important bug
outstanding, IMHO.
>
> But, please, let me argue again about the notes. They are VERY
> verbose, they give technical details about the implementation which
> mostly ALL other packages in Debian put in README.Debian to avoid
> interrupting installations.
>
OK, looking over this again, how about this:
README.Debian:
atlas3/ldlp
atlas3/nfs
Low priority notes and README.Debian:
atlas3/blas_lapack
atlas3-foo/foo_extensions
Medium(or high ?) priority notes:
atlas3-foo/no_foo
The idea is that the user should be made aware in some fashion at
runtime that certain libraries previously installed have been
transparently overriden. The mechanism is akin to a diversion, which,
if memory serves, does appear on the installation screen somewhere.
The user should be made especially aware that certain performance they
think they should be getting is not available, and in effect the
package is 'dead code'.
> But, moreover, as they use low priority, they are indeed NOT seen by
> probably 90 or 95% of your users. So, as such, they *are* useless. I
> suggest that you go through the details of the rationale given in the
> recent bug reports I sent: I couldn't develop better.
>
If all low priority notes are useless, then why does this priority
exist? Should they be medium, then?
> Please taken into account that the current maintainer of debconf (Joey
> Hess) himself thinks that low and medium priority notes fall under a
> contradiction and considers completely removing the support for notes.
>
I don't get this at all. Perhaps there should be but one priority
setting, but how else should a package be sure that the user sees
something? Can you imagine the spurious bug reports that should be
eliminated by well placed notes in a case like this?
> My proposal is to remove all the notes but the ones that are displayed
> when certain conditions on CPU's are not met.
>
> For instance:
>
> #!/bin/sh
>
> TDNCOMP=y
>
> . /usr/share/debconf/confmodule
>
> if ! grep ^flags /proc/cpuinfo | grep -q 3dnow; then
> db_input high atlas3-3dnow/no_3dnow
> fi
> db_go
>
>
> These "no_*" notes become *error* templates type which is perfectly
> acceptable.
>
OK, an error template is likely better -- is this new? How does it
differ from a high priority note?
>
> Finally, please also note that some of the templates are not listed in
> debian/po/POTFILES.in which makes them untranslatable and please also
> remember about the bug suggesting to run debconf-updatepo in the clean
> target.
>
OK.
> I still intent to work on a patch combining all these proposals, which
> should reduce the ammount of debconf stuff in atlas3 to an acceptable
> level. The corresponding patch will be sent to the BTS in #235732.
>
Thank you! Feedback on the above also most appreciated.
Take care,
>
--
Camm Maguire camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah
Reply to: