This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: isnan for ppc def help
- To: "Scully, Maureen" <maureen dot scully at lmco dot com>
- Subject: Re: isnan for ppc def help
- From: Matthew R Wette <mwette at grover dot jpl dot nasa dot gov>
- Date: Wed, 01 Mar 2000 09:21:37 -0800
- cc: "'Matthew dot R dot Wette at jpl dot nasa dot gov'" <Matthew dot R dot Wette at jpl dot nasa dot gov>, "'votava at fnal dot gov'" <votava at fnal dot gov>, "'nahn at fnal dot gov'" <nahn at fnal dot gov>, "'mayer at ik1 dot fzk dot de'" <mayer at ik1 dot fzk dot de>, "'gcc at gcc dot gnu dot org'" <gcc at gcc dot gnu dot org>
"Scully, Maureen" writes ...
> VxWorks user -
> Do you have any suggustions on how is can resolve the undefined
> symbol, isnan for the ppc?
> Some background info:
> I am using Tornado for VxWorks & Seawind for X/Motif/OpenGL support
> for VxWorks running on a PPC - power PC. Found your knowledgeable responses
> on a bulletin board when I search.
> I think that someone else must had the same problem or can advise me on
> porting software to the VxWorks/PPC environment. VxWorks/PPC Tornados
> Windshell outputs: undefined symbols: isnan, equal, ent, directory,
> valid_path, and remove_dir... These symbols are prototyped with a #ifdef
> unix and #ifdef mac differently. In these header files need to define a
> similar definition #ifdef VXWORKS for the PPC. isnan wrapper has the highest
> concern.
> I not sure don't understand why isnan differs in MAC and Unix defs. How then
> does VxWorks needs to do to resolve these symbols. Tornado's C++ compiler
> does not seem to resolve after I tried similar definitions during build
> because later during run time, spawning the loaded executable this problem
> still shows up. The NFS, network file system should help in resolving the
> variations in mkdir, chdir, rmdir. Not the comparable to isnan, valid_path,
> directory, ent, and equal under VxWorks.
> I think I need to load these symbol definitions in a *.o file, but I don
> understand them.
> I was hoping someone could share their insight into this symbol issue?
I find it odd that WRS left this out. Anyway, probably the easiest
thing to do is grab newlib from ftp://sourceware.cygnus.com/pub.
and grab the routine from there (for ppc!). I'm pretty sure it's
there. All that the routine has to do is test a bit in the fp status
register.
Matt