This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Add Android related compilation options and multilib options
On Tue, 16 Jun 2009, Doug Kwan (Ãö®¶¼w) wrote:
> I agree with you that a more flexible gcc is a good thing. However,
> to make that work, we need to build run-time libraries for each of the
> dynamic targets gcc is configured to support. Using multilibs may not
> be the best approach though.
My model for multi-target GCC is that runtime libraries would be
configured separately for each target; where config-ml.in now arranges for
them to be configured for each multilib, it would work over (target,
multilib) pairs, so from the point of view of each execution of each
library's configure script only one target and multilib would be relevant,
although the gcc/ directory configuration would need to handle multiple
targets (however, libgcc configuration would cease to copy information
from the gcc directory, and use of tm.h headers would be eliminated in
favour of the target structure and similar structures for the driver
etc.). Similarly, the testsuite would know the target name applicable to
the particular multilib being tested (different targets would need
different site.exp).
This does not answer the question of which configurations count as
different enough to have different triplets in this model rather than just
different multilibs for the same triplet.
--
Joseph S. Myers
joseph@codesourcery.com