Bizarre test for main in libm

[Loren James Rittle]

In article <200305011038.h41AcJm15772@pc960.cambridge.arm.com> you write:
> libstdc++-v3/aclocal.m4 contains the following test:
>
> dnl GLIBCPP_CHECK_COMPLEX_MATH_SUPPORT
> AC_DEFUN(GLIBCPP_CHECK_COMPLEX_MATH_SUPPORT, [
>  dnl Check for complex versions of math functions of platform.
>  AC_CHECK_LIB(m, main)
>
> Now this test for main in libm is broken as best I can see, since it is 
> expanded into
>
> int main() {
>  main()
> ; return 0; }
>
> Which of course always compiles and links without error (since it's 
> recursive).
>
> Does anyone know why that test is there in the first place?

I didn't know the answer off hand.  I'll take a stab with how I did it.

(First, realize that acinclude.m4 begets aclocal.m4 .)

; ccvs annotate libstdc++-v3/acinclude.m4 |grep 'm, main'
Annotations for libstdc++-v3/acinclude.m4
***************
1.77         (mmitchel 30-Oct-00):   AC_CHECK_LIB(m, main)

(Consult gcc-patches circa Oct 2000.)

http://gcc.gnu.org/ml/gcc-patches/2000-10/msg01007.html

Mark, mentioned that the patch was sort-a-ugly.  Are there any systems
where there is no -lm?  The above test would not link if -lm was not
found.  Correct?

Regards,
Loren