[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Rethinking Qt headers (should the header packages be recombined?)



On Thu, Feb 27, 2003 at 01:25:43PM +0100, Simon Richter wrote:
> 
> You are still assuming that the user is able to install software and
> doesn't have to grovel to her admin in order to get libqt3-dev installed
> only to find out that she also needs the -compat headers. That was my
> point.

Flip answer #1: don't use broken software.

> It doesn't suit the needs of people who want to install a Qt based
> theorem prover on the lab machines in the uni and have to run to their
> admins twice asking for a package to be installed on the next nightly
> maintenance run.

Semi-flip, semi-useful answer #2:

Download .deb, install .deb using --root into a user-available directory,
and make liberal use of -I to point the compiler at the 'local' copy, until
the admin has time to install a global one.

In fact, this works even if the admin never installed libqt3-dev, either.

The entire situation is less likely to occur if the package Recommends the
libqt3-compat-headers, which I think is a reasonable option (and I believe
was proposed by Ralf further up the thread).

Compat headers *are* deprecated, and not having them show up on a build
daemon is probbaly a very good thing at this point. Recommends means that
something like, oh, 95% of admins will install them, since it usually takes
some active effort to *not* do so using the UIs?

Any admin that goes to that effort probably has a reason for it, and that
is between them and their users. 'Normal' users of a home machine are
unlikely to get bitten by this, using Recommends. Developers are expected
to keep up with such events, especially if they're maintaining QT3-based
packages.
-- 
Joel Baker <fenton@debian.org>

Attachment: pgpyrOJ_Lw0PD.pgp
Description: PGP signature


Reply to: