[RFC][AArch64] Support in-tree building with libgloss and linker scripts (specs files)

Jeff Johnston jjohnstn@redhat.com
Wed Jan 29 17:13:00 GMT 2014

I'm ok with it since there is a work-around if needed and it is in an
architecture-specific directory.  So, go ahead.

-- Jeff J.

----- Original Message -----
From: "Corinna Vinschen" <vinschen@redhat.com>
To: newlib@sourceware.org
Sent: Wednesday, January 29, 2014 8:30:55 AM
Subject: Re: [RFC][AArch64] Support in-tree building with libgloss and linker scripts (specs files)

On Jan 29 13:09, Marcus Shawcroft wrote:
> On 24 January 2014 14:46, Richard Earnshaw <rearnsha@arm.com> wrote:
> > When using libgloss and linker scripts unified source tree builds
> > (gcc+newlib+binutils) fail when building the libraries because the
> > linker scripts have not been installed at that time.
> >
> > A suitable solution to this is to make copies of the scripts in the
> > build area, which the compiler is configured to search during the build.
> >  Then using LDFLAGS_FOR_TARGET=-specs=<spec-file> flag will correctly
> > pick up the appropriate libgloss specs file.
> >
> > I've marked this as an RFC as I'm not sure if the pattern-based rule is
> > GNUmake specific, and if it is, whether this is not appropriate for
> > newlib -- it could be written out long-hand, but this rule makes the
> > construct trivial to port to other libgloss targets.
> >
> > This was the only patch I needed to make an aarch64-elf unified tree
> > build and run the gcc testsuite.
> >
> > Thoughts please?
> Installing the scripts in the build area makes sense to me, but I
> think we need an ack from Jeff or Corinna w.r.t the use of GNU make
> extensions.

Personally I don't care.  If it's ok with Jeff, it's ok with me.


Corinna Vinschen
Cygwin Maintainer
Red Hat

More information about the Newlib mailing list