This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH][LTO] [1/3] Fix bootstrap with -flto
On Wed, Aug 19, 2009 at 09:12, Richard Guenther<rguenther@suse.de> wrote:
> *** 1108,1113 ****
> --- 1108,1117 ----
> Â Â/* Create the basic integer types. Â*/
> Â Âbuild_common_tree_nodes (flag_signed_char, /*signed_sizetype=*/false);
>
> + Â /* Share char_type_node with whatever is appropriate for this unit. Â*/
> + Â char_type_node
> + Â Â = flag_signed_char ? signed_char_type_node : unsigned_char_type_node;
> +
We cannot assume flag_signed_char will be present in lto1's command
line. At most, make it a flag that gets saved in the options section
of the bytecode files. But this will still fail if two TUs were
compiled with different values of flag_signed_char.
If flag_signed_char is completely resolved by each front end, then we
shouldn't have this issue. By completely resolved I mean that every
FE should emit one of signed_char_type_node or
unsigned_char_type_node. We should never see char_type_node on a
gimple stream.
Diego.