Le jeudi 06 décembre 2007 Ã* 16:23 -0500, Adam Jackson a écrit :
> The bdftruncate script provided by xorg-x11-font-utils drags in a
> dependency on perl. As far as I can tell, nothing in the distro uses
> it. The font-utils package is required by the various bitmap font
> packages, as it provides mkfontdir and mkfontscale.

Actually, I wrote last week I'd like font packages not to depend on
anything, be it "just" xorg-x11-font-utils of full perl. Pre-generating
fonts-* at build time is easy. Having dynamic fontconfig-like generation
less so but is not overly difficult either. Both would reduce
considerably the problems bdftruncate creates.

Since you seem to care about it do you want to take over core fonts
packaging guidelines maintenance? (I want a pony!)

> I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> unless someone has a compelling reason not to.

Nothing compelling, though it's a shame to drop functionalities for
packages which need a good dependency makeover anyway.

--
Nicolas Mailhot
--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

The bdftruncate script provided by xorg-x11-font-utils drags in a
dependency on perl. As far as I can tell, nothing in the distro uses
it. The font-utils package is required by the various bitmap font
packages, as it provides mkfontdir and mkfontscale.

I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
unless someone has a compelling reason not to.

- ajax

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

On Thu, 2007-12-06 at 22:09 +0100, Nicolas Mailhot wrote:
> Le jeudi 06 décembre 2007 Ã* 16:23 -0500, Adam Jackson a écrit :
> > The bdftruncate script provided by xorg-x11-font-utils drags in a
> > dependency on perl. As far as I can tell, nothing in the distro uses
> > it. The font-utils package is required by the various bitmap font
> > packages, as it provides mkfontdir and mkfontscale.
>
> Actually, I wrote last week I'd like font packages not to depend on
> anything, be it "just" xorg-x11-font-utils of full perl. Pre-generating
> fonts-* at build time is easy. Having dynamic fontconfig-like generation
> less so but is not overly difficult either. Both would reduce
> considerably the problems bdftruncate creates.

I agree that we should be able to package core fonts "up front" in the
sense of not needing processing in scriptlets, but I am utterly
unmotivated to make that happen.

> Since you seem to care about it do you want to take over core fonts
> packaging guidelines maintenance? (I want a pony!)

Not even a little bit. I don't care about core fonts working in the
slightest. My interest here is just on removing a rather bizarre arc in
the dep graph.

- ajax

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

Adam Jackson <ajackson <at> redhat.com> writes:
> I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> unless someone has a compelling reason not to.

What's the problem with depending on Perl? It's a useful scripting language,
used in a lot of packages, thus so many things require it that you'll be
hard-pressed to find a working system without Perl, especially an X11-enabled
one. For example, kdelibs requires Perl (it's used for things like
kconf_update), so any KDE system will always have Perl anyway.

Kevin Kofler

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

On Fri, 2007-12-07 at 06:12 +0000, Kevin Kofler wrote:
> Adam Jackson <ajackson <at> redhat.com> writes:
> > I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> > unless someone has a compelling reason not to.
>
> What's the problem with depending on Perl? It's a useful scripting language,
> used in a lot of packages, thus so many things require it that you'll be
> hard-pressed to find a working system without Perl, especially an X11-enabled
> one. For example, kdelibs requires Perl (it's used for things like
> kconf_update), so any KDE system will always have Perl anyway.

Perl is 10MB compressed and 30MB uncompressed. On a system like OLPC
that matters a lot. I remember when I worked on OLPC I did a lot of work
with danpb fixing up Fedora to remove such non-sense dependencies. We
need to care about things like these if we want Fedora to matter in
embedded and virt environments.

David


--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

On Fri, 2007-12-07 at 06:12 +0000, Kevin Kofler wrote:
> Adam Jackson <ajackson <at> redhat.com> writes:
> > I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> > unless someone has a compelling reason not to.
>
> What's the problem with depending on Perl? It's a useful scripting language,
> used in a lot of packages, thus so many things require it that you'll be
> hard-pressed to find a working system without Perl, especially an X11-enabled
> one. For example, kdelibs requires Perl (it's used for things like
> kconf_update), so any KDE system will always have Perl anyway.

Perl is 10MB compressed and 30MB uncompressed. On a system like OLPC
that matters a lot. I remember when I worked on OLPC I did a lot of work
with danpb fixing up Fedora to remove such non-sense dependencies. We
need to care about things like these if we want Fedora to matter in
embedded and virt environments.

David


--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

On Fri, 2007-12-07 at 10:08 -0500, Adam Jackson wrote:
> On Fri, 2007-12-07 at 06:12 +0000, Kevin Kofler wrote:
> > Adam Jackson <ajackson <at> redhat.com> writes:
> > > I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> > > unless someone has a compelling reason not to.
> >
> > What's the problem with depending on Perl? It's a useful scripting language,
> > used in a lot of packages, thus so many things require it that you'll be
> > hard-pressed to find a working system without Perl, especially an X11-enabled
> > one. For example, kdelibs requires Perl (it's used for things like
> > kconf_update), so any KDE system will always have Perl anyway.
>
> I'm not arguing against perl qua perl. I am arguing against needless
> arcs in the dep graph. This one package has a perl dependency solely
> because it provides a single perl script that no one is ever likely to
> use.
>
> Sure, it's just one dep, but every dep counts. The complexity of the
> graph directly affects runtime performance for rpm and yum, and the
> ability to use trimmed subsets of Fedora for custom purposes. The graph
> is important. We argue so much for correcness in initial packaging. We
> should be correct at the system level too.

And this is exactly the issue we faced with more constrained platforms
like OLPC; we had to hack that perl dep out of the X package because
_nothing_ else in the platform depended on perl, we didn't want to ship
perl, and perl takes up 20MB of space that we didn't have. Packagers
need to be more aggressive about splitting out deps from packages if
they want to be useful on targeted devices like the OLPC, the Classmate,
the eepc, and Fedora ARM.

Dan


--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

On Fri, 2007-12-07 at 06:12 +0000, Kevin Kofler wrote:
> Adam Jackson <ajackson <at> redhat.com> writes:
> > I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> > unless someone has a compelling reason not to.
>
> What's the problem with depending on Perl? It's a useful scripting language,
> used in a lot of packages, thus so many things require it that you'll be
> hard-pressed to find a working system without Perl, especially an X11-enabled
> one. For example, kdelibs requires Perl (it's used for things like
> kconf_update), so any KDE system will always have Perl anyway.

I'm not arguing against perl qua perl. I am arguing against needless
arcs in the dep graph. This one package has a perl dependency solely
because it provides a single perl script that no one is ever likely to
use.

Sure, it's just one dep, but every dep counts. The complexity of the
graph directly affects runtime performance for rpm and yum, and the
ability to use trimmed subsets of Fedora for custom purposes. The graph
is important. We argue so much for correcness in initial packaging. We
should be correct at the system level too.

- ajax

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

On Fri, 2007-12-07 at 10:08 -0500, Adam Jackson wrote:
> On Fri, 2007-12-07 at 06:12 +0000, Kevin Kofler wrote:
> > Adam Jackson <ajackson <at> redhat.com> writes:
> > > I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> > > unless someone has a compelling reason not to.
> >
> > What's the problem with depending on Perl? It's a useful scripting language,
> > used in a lot of packages, thus so many things require it that you'll be
> > hard-pressed to find a working system without Perl, especially an X11-enabled
> > one. For example, kdelibs requires Perl (it's used for things like
> > kconf_update), so any KDE system will always have Perl anyway.
>
> I'm not arguing against perl qua perl. I am arguing against needless
> arcs in the dep graph. This one package has a perl dependency solely
> because it provides a single perl script that no one is ever likely to
> use.
>
> Sure, it's just one dep, but every dep counts. The complexity of the
> graph directly affects runtime performance for rpm and yum, and the
> ability to use trimmed subsets of Fedora for custom purposes. The graph
> is important. We argue so much for correcness in initial packaging. We
> should be correct at the system level too.

Rather than dropping bdftruncate entirely, perhaps simply put it in its
own subpackage? You wouldn't need it to Require anything (the perl deps
will be autogenerated), nor would xorg-x11-font-utils need to Require
the bdftruncate subpackage, effectively solving the problem without
dropping (semi) useful code.

%package -n bdftruncate
Summary: Generate truncated BDF font from ISO 10646-1 encoded BDF font
Group: Applications/System

%description -n bdftruncate
bdftruncate allows one to generate from an ISO10646-1 encoded BDF font
other ISO10646-1 BDF fonts in which all characters above a threshold
code value are stored unencoded. This is often desirable because the
Xlib API and X11 protocol data structures used for representing font
metric information are extremely inefficient when handling sparsely
populated fonts.

...

%files -n bdftruncate
%defattr(-,root,root,-)
%{_bindir}/bdftruncate
%{_mandir}/bdftruncate.*

~spot

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

On Fri, 2007-12-07 at 10:08 -0500, Adam Jackson wrote:
> On Fri, 2007-12-07 at 06:12 +0000, Kevin Kofler wrote:
> > Adam Jackson <ajackson <at> redhat.com> writes:
> > > I plan to drop bdftruncate from font-utils to get rid of the perl bloat,
> > > unless someone has a compelling reason not to.
> >
> > What's the problem with depending on Perl? It's a useful scripting language,
> > used in a lot of packages, thus so many things require it that you'll be
> > hard-pressed to find a working system without Perl, especially an X11-enabled
> > one. For example, kdelibs requires Perl (it's used for things like
> > kconf_update), so any KDE system will always have Perl anyway.
>
> I'm not arguing against perl qua perl. I am arguing against needless
> arcs in the dep graph. This one package has a perl dependency solely
> because it provides a single perl script that no one is ever likely to
> use.
>
> Sure, it's just one dep, but every dep counts. The complexity of the
> graph directly affects runtime performance for rpm and yum, and the
> ability to use trimmed subsets of Fedora for custom purposes. The graph
> is important. We argue so much for correcness in initial packaging. We
> should be correct at the system level too.

Rather than dropping bdftruncate entirely, perhaps simply put it in its
own subpackage? You wouldn't need it to Require anything (the perl deps
will be autogenerated), nor would xorg-x11-font-utils need to Require
the bdftruncate subpackage, effectively solving the problem without
dropping (semi) useful code.

%package -n bdftruncate
Summary: Generate truncated BDF font from ISO 10646-1 encoded BDF font
Group: Applications/System

%description -n bdftruncate
bdftruncate allows one to generate from an ISO10646-1 encoded BDF font
other ISO10646-1 BDF fonts in which all characters above a threshold
code value are stored unencoded. This is often desirable because the
Xlib API and X11 protocol data structures used for representing font
metric information are extremely inefficient when handling sparsely
populated fonts.

...

%files -n bdftruncate
%defattr(-,root,root,-)
%{_bindir}/bdftruncate
%{_mandir}/bdftruncate.*

~spot

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list