gcc 2.95.2 bug: complex numbers and conditional op's don't mix

[Gabriel Dos Reis]

"Steven G. Johnson" <stevenj@superjdj.mit.edu> writes:

| Dear gcc folks,
| 
| I get an error "type mismatch in conditional expression" in gcc 2.95.2 (on
| Linux/x86) when compiling the following trivial and valid program.
| 
| The catch is that the program uses complex number types; it works fine if
| I use e.g. "double" instead, so this seems to be a bug in the complex
| number support.
| 
| Don't hesitate to email me if you need any other information.  I would
| also appreciate it if someone would try it in the latest CVS version of
| gcc and post whether they can reproduce the bug.  Thanks!

Current CVS exhibites the same behaviour.  The patch below fixes the
problem.

Bootstrap and tested on i686-pc-linux-gnu.

OK to commit?

I don't think the C++ front-end should be modified to accept the
construct. 

-- Gaby
CodeSourcery, LLC                             http://www.codesourcery.com


2000-06-04  Gabriel Dos Reis  <gdr@codesourcery.com>

	* c-typeck.c (build_conditional_expr): Handle complex data types.

Index: c-typeck.c
===================================================================
RCS file: /cvs/gcc/egcs/gcc/c-typeck.c,v
retrieving revision 1.67
diff -p -r1.67 c-typeck.c
*** c-typeck.c	2000/05/31 18:36:02	1.67
--- c-typeck.c	2000/06/04 21:31:30
*************** build_conditional_expr (ifexp, op1, op2)
*** 3359,3366 ****
        else
  	result_type = TYPE_MAIN_VARIANT (type1);
      }
!   else if ((code1 == INTEGER_TYPE || code1 == REAL_TYPE)
!            && (code2 == INTEGER_TYPE || code2 == REAL_TYPE))
      {
        result_type = common_type (type1, type2);
  
--- 3359,3368 ----
        else
  	result_type = TYPE_MAIN_VARIANT (type1);
      }
!   else if ((code1 == INTEGER_TYPE || code1 == REAL_TYPE
!             || code1 == COMPLEX_TYPE)
!            && (code2 == INTEGER_TYPE || code2 == REAL_TYPE
!                || code2 == COMPLEX_TYPE))
      {
        result_type = common_type (type1, type2);