[Fedora-packaging] Fedora Packaging Committee Meeting (Tuesday July 22)

[Richard W.M. Jones]

On Tue, Jul 22, 2008 at 08:06:59AM -0400, Jesse Keating wrote:
> On Tue, 2008-07-22 at 10:29 +0100, Daniel P. Berrange wrote:
> > What is the board's rationale for putting MinGW packages in a separate
> > repository, when other cross-compiler toolchain (eg ARM) are in the main
> > Fedora repository. Seems to me like we're penalizing MinGW  just
> > because it happens to be related to Windows, even though MinGW's code
> > is still just as open source as anything else in our repos.
> 
> Actually I think the prevailing thought that the Board has (although
> it's up to FESCo to really nail it down) is that the mingw tools
> themselves are absolutely suitable for Fedora.  The libraries compiled
> against it for windows use are what should be in another repo.

[I'm going to prepare something more detailed, hopefully integrating
efforts with the cross-compiler folks, but just on these two points ...]

If we ship only the four base packages (mingw-gcc, mingw-binutils,
mingw-w32api and mingw-runtime) then the only software that can be
compiled is software which doesn't use any libraries.  That's pretty
restrictive.

To compile, for example, libvirt, one needs six other libraries.  As
with Linux, you need the library around (foo-0.dll) in order to link.
Anyone compiling libvirt would need to download the source for each of
these six libraries and './configure --host=i686-pc-mingw32 ; make ;
make install' before they could start on libvirt, and of course it
isn't really that simple since those libraries don't all just
cross-compile without needing tweaks and patches.  Tweaks and patches
are what spec files are for.  This is why we'd like to ship
pre-compiled DLLs (only) of those six libs.

I think people have somehow got the impression we want to (a) ship
FIREFOX.EXE and/or (b) cross-compile every library in Fedora.  I'd
like to say that (a) is not our intention, ever, and (b) isn't even
technically possible, nevermind that it is completely undesirable.

> My personal opinion is that if you're going to need to munge spec files
> in order to produce packages built against mingw, those munges need to
> be done outside our cvs repo as well.

There are two ways that we've proposed that one could build
'mingw-gnutls'.  One is as a completely separate package, another is
as a subpackage of the ordinary gnutls.  I investigated and built
packages both ways (see links below) just to see what was technically
feasible.  It turns out that both methods are *technically* feasible.
Which is better from technical, organizational or political points of
view is a completely different question.

http://hg.et.redhat.com/misc/fedora-mingw--devel/?cmd=manifest;manifest=91a808c59de63589367c7bd9750da1fca342c529;path=/gnutls/
http://hg.et.redhat.com/misc/fedora-mingw--devel/?cmd=manifest;manifest=91a808c59de63589367c7bd9750da1fca342c529;path=/gnutls-fragment/

Rich.

-- 
Richard Jones, Emerging Technologies, Red Hat  http://et.redhat.com/~rjones
virt-p2v converts physical machines to virtual machines.  Boot with a
live CD or over the network (PXE) and turn machines into Xen guests.
http://et.redhat.com/~rjones/virt-p2v