Re: [Gcl-devel] Re: GCL netbsd/alpha

[Lord Isildur]

ive experienced this with numerous versions; it might be a gcc-on-alpha
thing. i will see what i can do about a newer one... one thing i noticed
for the first time just in the past couple of days is -fshort-double.
i wonder if this does anything. adding -mfp-regs -mno-soft-float
might also do something.

isildur



On Mon, 13 Oct 2003, Camm Maguire wrote:

> Greetings!  I see.  You know, I think your C compiler is very old,
> perhaps even experimental?
>
>
> Using builtin specs.
> gcc version egcs-2.91.60 19981201 (egcs-1.1.1 release)
> $
>
> Can you upgrade the compiler to something more recent, e.g. 3.3?  This
> kind of breakage is quite bad, IMHO, and perhaps while possible to
> work around, would appear unnecessary.
>
> Take care,
>
> Lord Isildur <address@hidden> writes:
>
> > hello,
> > i have experienced weird floating point errors using doubles and gcc from
> > time to time, enough that i generallyuse either long doubles or just plain
> > floats.. just one of the odd things ive gotten used to over the years.
> > (another is the interesting phenomenon that malloc cannot or will not give 
> > out
> > more than 64 megs out of the same stack frame..wrap a function call around
> > malloc, and it will give all the core you want- though not in chunks greater
> > than 64m- but call malloc out of the same frame, once or more than once,
> > and youll never get more than 64m successfully.. some temporary structure i
> > suppose, which malloc uses, and is sort of broken.. :)
> >
> > isildur
> >
> >
> > On Sat, 11 Oct 2003, Camm Maguire wrote:
> >
> > > greetings!
> > >
> > > Some weird floating point errors.  compiled with -mieee, which is
> > > required on debian alpha, but the following persists:
> > >
> > >
> > > In gdb:
> > >
> > > Breakpoint 3, make_shortfloat (f=Erroneous arithmetic operation.
> > > ) at number.c:148
> > > 148     {
> > >
> > > The code:
> > >
> > > object
> > > make_shortfloat(double f)
> > > {
> > >   object x;
> > >
> > >   if (f == (shortfloat)0.0)
> > >           return(shortfloat_zero);
> > >   x = alloc_object(t_shortfloat);
> > >   sf(x) = (shortfloat)f;
> > >   return(x);
> > > }
> > >
> > > The call:
> > >
> > >             make_constant("LEAST-POSITIVE-SHORT-FLOAT",
> > > 1395                          make_shortfloat(smallest_float));
> > > 1396
> > >
> > > can't a float be cast to a double in a function call?  Do you know of
> > > a fix?
> > >
> > > Take care,
> > >
> > >
> > > Lord Isildur <address@hidden> writes:
> > >
> > > > On Fri, 10 Oct 2003, Camm Maguire wrote:
> > > > > missing x libraries -- cannot compile xgcl
> > > >
> > > > also, the X libraries in netbsd are in /usr/lib/X11 . It should be 
> > > > pretty much
> > > > clean Xfree, from there on.
> > > >
> > > > Isildur
> > > >
> > > >
> > > > _______________________________________________
> > > > Gcl-devel mailing list
> > > > address@hidden
> > > > http://mail.gnu.org/mailman/listinfo/gcl-devel
> > > >
> > > >
> > > >
> > >
> > > --
> > > Camm Maguire                                              address@hidden
> > > ==========================================================================
> > > "The earth is but one country, and mankind its citizens."  --  Baha'u'llah
> > >
> > >
> >
> >
> > _______________________________________________
> > Gcl-devel mailing list
> > address@hidden
> > http://mail.gnu.org/mailman/listinfo/gcl-devel
> >
> >
> >
>
> --
> Camm Maguire                                          address@hidden
> ==========================================================================
> "The earth is but one country, and mankind its citizens."  --  Baha'u'llah
>
>