Hi Nathan,
On 7/22/07, Nathan Sidwell <nathan@codesourcery.com> wrote:
Why not just have cp_parser_asm_operand_list return error_mark_node in the face
of any bogus operand? (with suitable operand eating in that case)?
The attached patch implements this suggestion by modifying
cp_parser_asm_operand_list to track whether any of the operands are
invalid and to return error_mark_node in that case, adjusting the
checks in cp_parser_asm_definition accordingly.
Bootstrapped and regression tested with no new failures on
i686-pc-linux-gnu. Ok to apply?