This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Don't build abi_check for crosses
- From: Benjamin Kosnik <bkoz at redhat dot com>
- To: Daniel Jacobowitz <drow at mvista dot com>
- Cc: gcc-patches at gcc dot gnu dot org, aj at suse dot de, libstdc++ at gcc dot gnu dot org
- Date: Wed, 10 Sep 2003 04:50:46 -0500
- Subject: Re: Don't build abi_check for crosses
- Organization: Red Hat / Paris
- References: <20030908200206.GA29722@nevyn.them.org>
>This is necessary because automake can't build abi_check for the host.
>Right now it's listed as a target program, and will be linked to the newly
>built libstdc++; if $build != $host or --with-cross-host, we won't be able
>to run it even if we were able to build it. Ben, CC'd you because you
>removed the comment saying to build it native only; was there a reason?
FYI for the testsuite and testsuite directory at build time, CXX is g++,
not gcc. See scripts/testsuite_hooks.in. Thus, this should link, even
for crosses and other twisted permutations, and if it doesn't you will
be in for a world of pain at the next state of your build when you try
to test and or invoke g++. There are (were?) known problems and issues
with linking on mipsel in particular, echristo knows the deal, it's
something with weak linking. However....
Right now, considering the limitations of the ABI checking, I think the
'check-abi' rule only makes sense for native targets. Thus, I think your
patch is ok, and trust that you will resolve the rest of the mipsel
issues with linking separately.
best,
-benjamin
>2003-09-08 Daniel Jacobowitz <drow@mvista.com>
> Andreas Jaeger <aj@suse.de>
>
> PR libstdc++/12189
> * acinclude.m4 (GLIBCPP_CONFIGURE_TESTSUITE): Don't build
> abi_check if cross compiling.
> * aclocal.m4: Regenerated.
> * configure: Regenerated.