[Bug middle-end/35509] [4.3/4.4 Regression] builtin isinf() mismatch to compile-time substitution

[ubizjak at gmail dot com]

------- Comment #4 from ubizjak at gmail dot com  2008-03-08 15:51 -------
(In reply to comment #2)

> (Only i386 and s390 have target dependent expansions for isinf - what is
> their behavior here?)
> 
> Kaveh, you introduced the generic fallback, Andreas, what does s390 return
> here, Uros, how is the situation on i386?

When compiled on 32bit target with -std=c99, testcase aborts. For !c99, library
call is emitted and the test passes OK.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35509