lubuntu-desktop team mailing list archive

[Jonathan Marsden <jmarsden@xxxxxxxxxxx>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 02/18/2011 12:02 PM, Jean-Pierre Vidal Piesset wrote:

> So, according to you it doesn't affect that each time a package is 
> installed or uninstalled the system tries to find something and
> fails:

Correct... however, I do not think it's primarily my role to say what it
affects -- because I am not the person reporting the bug :)  What do
*you* think it affects?  That is what I am asking you about.

You are the one suggesting that you have found a bug in debconf;
therefore, I'm asking you for clear information about why displaying a
warning message is a bug -- what specific impact does it have for you
(the bug reporter) that makes you think it is a bug?

Personally, I have seen the same warning message in other (non-Lubuntu)
circumstances, and it has never been a problem for me.  Therefore, I
currently believe it to be normal standard expected "works-as-designed"
debconf behaviour, and I do not think it has a significant performance
impact.  But, I'm not the one reporting this as a bug :)

There may be ways to avoid the warning by reconfiguring debconf, have
you explored using

  sudo dpkg-reconfigure debconf

to suit your needs?  Or, if that does not help, starting upgrade-manager
using something like:

  DEBIAN_FRONTEND=dialog update-manager

or possibly even

  export DEBIAN_FRONTEND=dialog ; update-manager

to explicitly set the desired front end?  One or both ought to work to
avoid this warning, if I am guessing correctly :)

> this is a process that sure last less than a second... but Lubuntu
> is for old-spec hardware, so "here", "less than a second" can be a
> second for "other person" with older hardware (and multiplied for
> each package you're un/installing...).

OK.  Have you run strace to see how much "less than a second" checking
for the presence of libgnome-perl takes, or what system resources are
used to make that check and output the warning message?  Have you read
the debconf sources to see exactly what the code does to perform the
check, and estimated that what it does when it makes that check will
cause performance issues on older hardware?  Have you installed say 200
packages under GNOME, timing how long it takes, and then repeated the
same install of the same 200 packages under Lubuntu on the same
hardware, and found that Lubuntu is significantly slower doing that install?

If not, then what makes you think that it has a noticeable performance
impact for you -- you are the bug reporter; please provide the info that
makes you think this (checking for ability to use some Perl code from
libgnome2-perl, and then emitting a text warning in a debug window that
is usually hidden) is causing a problem for you, and so is a bug.

In case it helps you as you look into this, the code concerned is in
Debconf/FrontEnd/Gnome.pm and reads:

    eval q{
        use Gtk2;
        use Gnome2;
    };
    die "Unable to load Gnome -- is libgnome2-perl installed?\n" if $@;

So debconf is loading its Gnome front end module, which tests whether it
has the Perl stuff it needs, finds out it doesn't, and quits, causing
the front end selector to fall back to the Dialog front end.  I don't
see an obvious major performance issue there, but only some real world
testing will find out for sure.

I am suspicious that the "fault" here, if any, could be that
update-manager runs debconf specifying the gnome frontend by default.

Do you still see the issue if you invoke update-manager as

  DEBIAN_FRONTEND=dialog update-manager

> And so, everybody here knows that Lubuntu doesn't include Gnome...
> that's why a call for something gnome-called looks bad.

It's not even visible to 98% (a guess, I have no evidence what the
percentage really is) of users -- all the users who don't play with the
Details button when running update-manager.

> The process automatically falls back to the frontend "dialog"... i'm
> just saying that if it's possible, this should be done by default in
> Lubuntu.

I'd think that yes, it's *possible*.  Most things in software are
*possible* given enough time and work.  But that does not in any way
suggest that everything which is possible "should be done".

I think a more helpful question is not whether it is possible, but
whether it is worth doing the work to hide a warning that is already
mostly hidden :)

If you want to take this request further, I would suggest filing a bug
report in LaunchPad and providing full details there.

Thanks,

Jonathan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk1fIywACgkQUGfT4+mKBLKqXgCfdOQyiTRFzNIacEWnPkTWruap
MnYAnA4CYIAUy3eau0S5TDJT4ueByjbq
=w2zO
-----END PGP SIGNATURE-----