This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [Tru64 UNIX V5.1] gcc testsuite fixes

On 28-Feb-2001, Rainer Orth <ro@TechFak.Uni-Bielefeld.DE> wrote:
> In gcc.log, I find that compilation fails with a linker error:
> /usr/bin/ld:
> Warning: nan defined as GLOBAL DATA but is defined in a shared lib as a GLOBAL FUNC

That's only a warning.  Does the link really fail?

> Wed Feb 28 20:45:52 2001  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
> 	* gcc.c-torture/execute/ieee/fp-cmp-1.c: Renamed nan to dnan.
> 	* gcc.c-torture/execute/ieee/fp-cmp-3.c: Likewise.
> 	* gcc.c-torture/execute/ieee/fp-cmp-2.c: Renamed nan to fnan.

That shouldn't be necessary, at least if those tests are compiled with
`-std=c89'; in C89, `nan' is in the user's namespace, not the
implementation's, and so a user should be allowed to use that name.
If the implementation doesn't handle that case right, then it is a bug.

(Note that nan() is a standard library function that was introduced in

Fergus Henderson <>  |  "I have always known that the pursuit
                                    |  of excellence is a lethal habit"
WWW: <>  |     -- the last words of T. S. Garp.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]