[v10,01/10] rustc: new virtual package

The compiler for the Rust programming language is called rustc.

There is only one reference implementation for it, based on LLVM, from
the Rust project [1]. It can generate code for various architectures so
it can be labeled as a cross-compiler. But, as for GCC, building it
from source takes time.

So it would be sensible to have at least one package which provides it
as a pre-built version, fetched from the upstream project. Later another
package can be added, to build it from source code.

In addition to the compiler, the standard library for the host and/or
the target should also be fetched/built.

So, add a virtual package named rustc to enable support for multiple
providers.

Currently, only the host variant will be available to allow the user to
cross-compile Rust programs for the target.

[1] http://rust-lang.org

Signed-off-by: Eric Le Bihan <eric.le.bihan.dev@free.fr>
---
 package/Config.in.host       |  1 +
 package/rustc/Config.in.host | 31 +++++++++++++++++++++++++++++++
 package/rustc/rustc.mk       | 20 ++++++++++++++++++++
 3 files changed, 52 insertions(+)
 create mode 100644 package/rustc/Config.in.host
 create mode 100644 package/rustc/rustc.mk

Hello,

+Peter, Arnout, Ricardo in Cc.

On Sun,  4 Feb 2018 19:07:40 +0100, Eric Le Bihan wrote:

> diff --git a/package/rustc/rustc.mk b/package/rustc/rustc.mk
> new file mode 100644
> index 0000000000..7a951c0cfd
> --- /dev/null
> +++ b/package/rustc/rustc.mk
> @@ -0,0 +1,20 @@
> +################################################################################
> +#
> +# rustc
> +#
> +################################################################################
> +
> +RUSTC_ARCH = $(call qstrip,$(BR2_PACKAGE_HOST_RUSTC_ARCH))
> +RUSTC_ABI = $(call qstrip,$(BR2_PACKAGE_HOST_RUSTC_ABI))
> +
> +RUST_TARGET_NAME = $(RUSTC_ARCH)-unknown-linux-gnu$(RUSTC_ABI)
> +
> +ifeq ($(HOSTARCH),x86)
> +RUSTC_HOST_ARCH = i686
> +else
> +RUSTC_HOST_ARCH = $(HOSTARCH)
> +endif
> +
> +RUST_HOST_NAME = $(RUSTC_HOST_ARCH)-unknown-linux-gnu

check-package is not happy with RUST_TARGET_NAME and RUST_HOST_NAME
because they don't have a RUSTC_ prefix, even though they are defined
in rustc.mk:

./package/rustc/rustc.mk:10: possible typo: RUST_TARGET_NAME -> *RUSTC*
./package/rustc/rustc.mk:18: possible typo: RUST_HOST_NAME -> *RUSTC*

Would renaming them to RUSTC_TARGET_NAME and RUSTC_HOST_NAME make
sense ? Or do we need to keep RUST_TARGET_NAME/RUST_HOST_NAME and add
an exception in check-package ?

Thanks for your feedback,

Thomas

On 2018-02-08 23:52, Thomas Petazzoni wrote:
> Hello,
>
> +Peter, Arnout, Ricardo in Cc.
>
> On Sun,  4 Feb 2018 19:07:40 +0100, Eric Le Bihan wrote:
>
> > diff --git a/package/rustc/rustc.mk b/package/rustc/rustc.mk
> > new file mode 100644
> > index 0000000000..7a951c0cfd
> > --- /dev/null
> > +++ b/package/rustc/rustc.mk
> > @@ -0,0 +1,20 @@
> > +################################################################################
> > +#
> > +# rustc
> > +#
> > +################################################################################
> > +
> > +RUSTC_ARCH = $(call qstrip,$(BR2_PACKAGE_HOST_RUSTC_ARCH))
> > +RUSTC_ABI = $(call qstrip,$(BR2_PACKAGE_HOST_RUSTC_ABI))
> > +
> > +RUST_TARGET_NAME = $(RUSTC_ARCH)-unknown-linux-gnu$(RUSTC_ABI)
> > +
> > +ifeq ($(HOSTARCH),x86)
> > +RUSTC_HOST_ARCH = i686
> > +else
> > +RUSTC_HOST_ARCH = $(HOSTARCH)
> > +endif
> > +
> > +RUST_HOST_NAME = $(RUSTC_HOST_ARCH)-unknown-linux-gnu
>
> check-package is not happy with RUST_TARGET_NAME and RUST_HOST_NAME
> because they don't have a RUSTC_ prefix, even though they are defined
> in rustc.mk:
>
> ./package/rustc/rustc.mk:10: possible typo: RUST_TARGET_NAME -> *RUSTC*
> ./package/rustc/rustc.mk:18: possible typo: RUST_HOST_NAME -> *RUSTC*
>
> Would renaming them to RUSTC_TARGET_NAME and RUSTC_HOST_NAME make
> sense ? Or do we need to keep RUST_TARGET_NAME/RUST_HOST_NAME and add
> an exception in check-package ?

I do not see any problems using RUSTC_{HOST,TARGET}_NAME: it still caries
the same information as RUST_{HOST,TARGET}_NAME.

Regards,

--
ELB

Hello,

On Fri, 9 Feb 2018 09:49:06 +0100, Eric Le Bihan wrote:

> > check-package is not happy with RUST_TARGET_NAME and RUST_HOST_NAME
> > because they don't have a RUSTC_ prefix, even though they are defined
> > in rustc.mk:
> >
> > ./package/rustc/rustc.mk:10: possible typo: RUST_TARGET_NAME -> *RUSTC*
> > ./package/rustc/rustc.mk:18: possible typo: RUST_HOST_NAME -> *RUSTC*
> >
> > Would renaming them to RUSTC_TARGET_NAME and RUSTC_HOST_NAME make
> > sense ? Or do we need to keep RUST_TARGET_NAME/RUST_HOST_NAME and add
> > an exception in check-package ?  
> 
> I do not see any problems using RUSTC_{HOST,TARGET}_NAME: it still caries
> the same information as RUST_{HOST,TARGET}_NAME.

Question is: are these host and target names "Rust" related, or "Rust
compiler" related. If they are Rust related, the current RUST_ prefix
makes more sense. If they are specific to the Rust compiler, then
RUSTC_ would make more sense. If it isn't clear, then we can use RUSTC_
to avoid the problem :-)

If you are happy with RUSTC_, could you submit a patch against "master"
that does the renaming ?

Thanks!

Thomas

On 2018-02-09 09:56, Thomas Petazzoni wrote:
> Hello,
>
> On Fri, 9 Feb 2018 09:49:06 +0100, Eric Le Bihan wrote:
>
> > > check-package is not happy with RUST_TARGET_NAME and RUST_HOST_NAME
> > > because they don't have a RUSTC_ prefix, even though they are defined
> > > in rustc.mk:
> > >
> > > ./package/rustc/rustc.mk:10: possible typo: RUST_TARGET_NAME -> *RUSTC*
> > > ./package/rustc/rustc.mk:18: possible typo: RUST_HOST_NAME -> *RUSTC*
> > >
> > > Would renaming them to RUSTC_TARGET_NAME and RUSTC_HOST_NAME make
> > > sense ? Or do we need to keep RUST_TARGET_NAME/RUST_HOST_NAME and add
> > > an exception in check-package ?
> >
> > I do not see any problems using RUSTC_{HOST,TARGET}_NAME: it still caries
> > the same information as RUST_{HOST,TARGET}_NAME.
>
> Question is: are these host and target names "Rust" related, or "Rust
> compiler" related. If they are Rust related, the current RUST_ prefix
> makes more sense. If they are specific to the Rust compiler, then
> RUSTC_ would make more sense. If it isn't clear, then we can use RUSTC_
> to avoid the problem :-)

As RUSTC_{HOST,TARGET}_NAME is passed to the Rust compiler by the
--target option, it is "Rust compiler" related. So the renaming makes
sense.

> If you are happy with RUSTC_, could you submit a patch against "master"
> that does the renaming ?

Should I provide a global patch or a series of patches, one for each
affected package?

Regards,

--
ELB