Re: powerpc-eabism, strategy for finding crt0.o

[Geoff Keating <geoffk at geoffk dot org>]

> Cc: gcc-patches@gcc.gnu.org, geoffk@cygnus.com, newlib@sources.redhat.com
> From: Alexandre Oliva <aoliva@redhat.com>
> Organization: GCC Team, Red Hat
> Date: 12 Apr 2001 06:35:29 -0300
> User-Agent: Gnus/5.0808 (Gnus v5.8.8) XEmacs/21.1 (Cuyahoga Valley)
> 
> On Apr 12, 2001, Benjamin Kosnik <bkoz@redhat.com> wrote:
> 
> >> I suggest the following:
> >> 
> >> Move the case ${target} from libgloss/configure.in into a separate
> >> shell-script fragment, that sets some variable to the (list of)
> >> directories in which GCC should search for crt files and libraries.
> >> 
> >> Then, get the top-level configure.in to source this shell-script
> >> fragment and add -B$$r/$(TARGET_SUBDIR)/libgloss/$libglosssubdir(s)/
> >> to FLAGS_FOR_TARGET.
> 
> > This would be great. Then everything would "just work" with what is 
> > currently in CVS for libstdc++-v3.
>  
> Yep.  I suppose this new config fragment could also set any additional
> flags needed for link tests to succeed.  I don't think we need flags
> specific for the simulator, since I don't think many embedded targets
> have different triplets for simulator or board set-ups.  Presumably,
> we only need to be able to link, since, when cross compiling, we don't
> run configure tests.  So, simulator-specific flags shouldn't be
> necessary.  Right, Geoff?

No, the compiler will still need -msim to link correctly for the
simulator when configured as 'powerpc-eabi'.  It has to be told which
board it is linking for.

While we don't run tests in configure, we still need to run
dejagnu tests.

-- 
- Geoffrey Keating <geoffk@geoffk.org>