This is the mail archive of the
mailing list for the GCC project.
Re: [patch, spu] More flexible --with-sysroot to enable "common" PPU/SPU sysroot
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: joseph at codesourcery dot com (Joseph S. Myers)
- Cc: gcc-patches at gcc dot gnu dot org, Trevor_Smigiel at playstation dot sony dot com, Andrew_Pinski at playstation dot sony dot com
- Date: Mon, 17 Nov 2008 21:20:42 +0100 (CET)
- Subject: Re: [patch, spu] More flexible --with-sysroot to enable "common" PPU/SPU sysroot
Joseph Myers wrote:
> On Fri, 14 Nov 2008, Ulrich Weigand wrote:
> > I'd like to take this issue up again and propose a modified version of
> > the patch. This version changes the SPU back-end to simply define
> > the "native" install paths as "/include" and "/lib" -- just like the
> > GNU target does, for example.
> I don't think this is back-end-specific; I think such paths are
> appropriate for all bare-metal targets where there isn't really a "native"
> install. There's another use case for configuring such targets with
> sysroots: if the path where you are installing parts of the (relocatable)
> toolchain during building is a temporary path different from the
> configured prefix, you can get a sysrooted compiler to find headers and
> libraries in that temporary path using --with-build-sysroot, but there's
> no equivalent option for non-sysroot configurations, and right now using a
> sysroot configuration requires a usr -> . symlink while building the
I agree that other platforms might not want the hard-coded "/usr" as well.
However, whether this is the case for any specific platform does appear to
be an issue that needs to be selected per-platform.
I'm wondering how else the decision could be made except via the back-end
defines my patch is setting. Do you have any suggestions what to use
instead? A new configure.tgt setting, maybe?
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE