Fedora Board Recap 2008-JUL-15

On Fri, 2008-07-18 at 06:50 +0200, Ralf Corsepius wrote:
> I feel a lot of people don't want to comprehend that cross-toolchains
> are ordinary (Fedora-)native applications and do not need any special
> treatment besides the spots where they collide with current policies.

The cross-toolchain (mingw-gcc, mingw-binutils) are ordinary enough.
What is concerning is that:

1. Even though there is a practical standard for cross-compilers, no one
has yet written it down.

2. Some of the folks in the conversation (the libvirt folks) want to
build other (non-toolchain) packages with the mingw cross-toolchain.

~spot

_______________________________________________
fedora-advisory-board mailing list
fedora-advisory-board@redhat.com
http://www.redhat.com/mailman/listinfo/fedora-advisory-board

On Thu, 2008-07-17 at 19:57 -0800, Jeff Spaleta wrote:
> What I would prefer to do is make sure that I am at the FESCo meeting
> where this topic is introduced, so I can act as a resource to FESCo.
> We should make sure that spot is there as well I guess. If FESCo
> wants me to personally tap me to draft up my braindump on the
> packaging policy, by working with Richard Jones and the other
> developers who have initiated this.. outside of the public lists...I
> can do that, as long as I haven't burned too many bridges in the fab
> conversation leading up to this moment.

Please do so. Your work on the Spins policies was extremely appreciated
and I think you can make a good impact here as well.

josh

_______________________________________________
fedora-advisory-board mailing list
fedora-advisory-board@redhat.com
http://www.redhat.com/mailman/listinfo/fedora-advisory-board

On Fri, 2008-07-18 at 08:13 -0400, Tom "spot" Callaway wrote:
> On Fri, 2008-07-18 at 06:50 +0200, Ralf Corsepius wrote:
> > I feel a lot of people don't want to comprehend that cross-toolchains
> > are ordinary (Fedora-)native applications and do not need any special
> > treatment besides the spots where they collide with current policies.
>
> The cross-toolchain (mingw-gcc, mingw-binutils) are ordinary enough.
> What is concerning is that:
>
> 1. Even though there is a practical standard for cross-compilers, no one
> has yet written it down.
Which standard do you need?

There is only one point where GNU-cross-toolchains collide with our
current guidelines:

${exec_prefix}/<target-alias> is defacto practice in GNU-toolchains for
a long time (decades), but collides with the FHS. Unfortunately,
changing this would require a non-trivial amount of work.


Technically, there are many issues related to rpm and redhat-rpm-config,
mostly related to them lacking generality (e.g. not taking into account
foreign binary formats/debug-info generation not being able to
distinguish between target and host sources).


> 2. Some of the folks in the conversation (the libvirt folks) want to
> build other (non-toolchain) packages with the mingw cross-toolchain.
Well, building foreign binaries is the purpose of using a
cross-toolchain That said, wrt. MinGW, which to allow and which not
is a political issue. I would lean to not to shipping any but some
"hello world" or "FreeDOS"-class testcases.

However, even this isn't a new topic: We already have several
cross-toolchains (off-head, IIRC, the only GNU cross-toolchain is the
avr cross-toolchain, but there already are others) and several emulators
in Fedora.

Ralf



_______________________________________________
fedora-advisory-board mailing list
fedora-advisory-board@redhat.com
http://www.redhat.com/mailman/listinfo/fedora-advisory-board

On Fri, 2008-07-18 at 14:49 +0200, Ralf Corsepius wrote:
> > 1. Even though there is a practical standard for cross-compilers, no
> one
> > has yet written it down.
> Which standard do you need?
>
> There is only one point where GNU-cross-toolchains collide with our
> current guidelines:
>
> ${exec_prefix}/<target-alias> is defacto practice in GNU-toolchains
> for
> a long time (decades), but collides with the FHS. Unfortunately,
> changing this would require a non-trivial amount of work.

Okay, so documenting this, as well as the naming standard (for the
toolchain and the packages) is what I was referring to. Or to be blunt:

If I wanted to package a sparc cross-compiler, what would I need to know
to do it properly in Fedora? Our written guidelines right now don't
cover it at all (as far as I know).

And yes, I know we have other cross toolchains, but this is the first
time anyone has wanted to package items in Fedora which are built from
those cross toolchains. I agree with your approach of having nothing in
the main repository other than simple testcases in the toolchain
packages, but I also don't have a problem with permitting these packages
to live in a separated location.

~spot

_______________________________________________
fedora-advisory-board mailing list
fedora-advisory-board@redhat.com
http://www.redhat.com/mailman/listinfo/fedora-advisory-board

On Fri, 2008-07-18 at 08:58 -0400, Tom "spot" Callaway wrote:
> On Fri, 2008-07-18 at 14:49 +0200, Ralf Corsepius wrote:
> > > 1. Even though there is a practical standard for cross-compilers, no
> > one
> > > has yet written it down.
> > Which standard do you need?
> >
> > There is only one point where GNU-cross-toolchains collide with our
> > current guidelines:
> >
> > ${exec_prefix}/<target-alias> is defacto practice in GNU-toolchains
> > for
> > a long time (decades), but collides with the FHS. Unfortunately,
> > changing this would require a non-trivial amount of work.
>
> Okay, so documenting this, as well as the naming standard (for the
> toolchain and the packages) is what I was referring to. Or to be blunt:
>
> If I wanted to package a sparc cross-compiler, what would I need to know
> to do it properly in Fedora? Our written guidelines right now don't
> cover it at all (as far as I know).
>
> And yes, I know we have other cross toolchains, but this is the first
> time anyone has wanted to package items in Fedora which are built from
> those cross toolchains. I agree with your approach of having nothing in
> the main repository other than simple testcases in the toolchain
> packages, but I also don't have a problem with permitting these packages
> to live in a separated location.

And just to make another piece of our Board discussion more transparent:
Independent of implementation, a limit to the space set aside for these
products would provide a motivation for people *not* to want to go to
the extreme case Jef was concerned about -- cross compiling "every
library" in Fedora.

--
Paul W. Frields
gpg fingerprint: 3DA6 A0AC 6D58 FEC4 0233 5906 ACDB C937 BD11 3717
http://paul.frields.org/ - - http://pfrields.fedorapeople.org/
irc.freenode.net: stickster @ #fedora-docs, #fedora-devel, #fredlug
_______________________________________________
fedora-advisory-board mailing list
fedora-advisory-board@redhat.com
http://www.redhat.com/mailman/listinfo/fedora-advisory-board