((bool) -0.5) mishandled by Sun C 5.9 alpha, which causes configure failure

Question

In <http://lists.gnu.org/archive/html/bug-autoconf/2006-05/msg00027.html>
Bruno Haible reports that Sun C 5.9 alpha for Linux/x86 mishandles the
following test case:
#include <stdbool.h>
char d[(bool) -0.5 == true ? 1 : -1];
This is strictly conforming C99 code, and GCC accepts it, but Sun C
5.9 alpha apparently rejects it with the diagnostic "integral constant
expression expected".
((bool) -0.5) is an integer constant expression, because bool is an
integer type (Bruno's analysis of this particular point is incorrect,
as you'll be able to see when I post my followup to his email). Hence
Sun C 5.9 alpha is buggy.
This bug causes 'configure' to conclude that Sun C 5.9's
implementation of 'bool' is inadequate, and programs typically use
their own (inferior) substitute for 'bool'.
Can you please fix this?
Thanks.
PS. One minor point of terminology. The C standard now uses the term
"integer constant expression", not "integral constant expression". (I
vaguely recall that "integral" was the old term.) So you should
update the wording in your diagnostics.

Answer 1

Here is a test case.
============================= boolcast.c =============================
#include <stdbool.h>
char ac[(int) 3.999 == 3 ? 1 : -1];
char am[((int) 3.999 == 3) * 2 - 1];
char dc[(bool) -0.5 == true ? 1 : -1];
char dm[((bool) -0.5 == true) * 2 - 1];
char fc[(_Bool) -0.0 == false ? 1 : -1];
char fm[((_Bool) -0.0 == false) * 2 - 1];======================================================================
$ c99 -Xa -c boolcast.c
"boolcast.c", line 6: integral constant expression expected
"boolcast.c", line 6: zero or negative subscript
"boolcast.c", line 7: warning: can not declare variably modified type at file scope
"boolcast.c", line 9: integral constant expression expected
"boolcast.c", line 9: zero or negative subscript
"boolcast.c", line 10: warning: can not declare variably modified type at file scopeThe argumentation is as follows:
- ISO C 99, section 6.2.5 paragraph 6, says that _Bool is an unsigned integer type.
- ISO C 99, section 6.2.5 paragraph 17, then implies that _Bool is an integer type.
- ISO C 99, section 6.6 paragraph 6, defines the term "integer constant
expression":
An integer constant expression shall have integer type and shall
only have operands that are integer constants, enumeration constants,
character constants, sizeof expressions whose results are integer
constants, and floating constants that are the immediate operands of
casts. Cast operators in an integer constant expression shall only
convert arithmetic types to integer types, except as part of an operand
to the sizeof operator.
This is practically relevant: These kinds of constructions are used by the
GNU autoconf AC_HEADER_STDBOOL macro.
Bruno