Re: Does libstdc++ v3 not work with Canadian crosses?

[Daniel Jacobowitz <dmj+ at andrew dot cmu dot edu>]

On Thu, Jun 28, 2001 at 06:04:06AM -0300, Alexandre Oliva wrote:
> On Jun 27, 2001, Daniel Jacobowitz <dmj+@andrew.cmu.edu> wrote:
> 
> >   # If Canadian cross, then don't pick up tools from the build
> >   # directory.
> >   if test x"$build" != x"$with_cross_host" && x"$build" != x"$target"; then
> 
> > There'd better be a "test" after that "&&" unless my installed shells
> > have diverged from everyone else's.
> 
> Yup, this is definitely wrong.  Especially given that, when you
> configure the top-level with a certain (build,host,target) triplet,
> target libraries such as libstdc++ are configured with
> (host,target,N/A) for (build,host,target), so testing for target in
> libstdc++'s configure is totally wrong.

Is that really true?

I configured the top level with:
  --build=i686-pc-linux-gnu
  --host=mipsel-hardhat-linux
  --target=mipsel-hardhat-linux

Libstdc++ was configured with:
  --build=i686-pc-linux-gnu
  --host=mipsel-hardhat-linux
  --with-target-subdir=mipsel-hardhat-linux

i.e. build is preserved.

Soon as I understand what's going on with this build a little better,
I'll submit a patch for Canadian crosses; I suspect something not long
before release broke it, but the released 3.0 definitely does not build
them properly.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer