[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 971209: netCDF on FreeBSD

> > If it's not too much trouble, grab
> > ftp://ftp.unidata.ucar.edu/pub/netcdf/netcdf-3.4atd.tar.Z
> > This is the release we are currently testing.
> > I think that there are some changes to nc_test
> > which give more meaningful diagnostics.
>
>
> I have grabbed, built and tested and the test gives a similar
> output. There are some values that are equal but when compared
> don't seem so.. (-128==-128.?)
>
> Anyway, there are others values that don't appear explicitly written in
> the output. I'll check them one by one, but it will take me some time, and
> I have some other more urgent tasks for the next month +/- After
> having checked them, I will contact you again to
> report the results of this detailed check.
> As a whole, the number of errors is less than with previous version, but I
> wouldn't reliably use this built unless I checked these problems one by
> one...
>
> If we get this version working under FreeBSD (reliably!), I will
> prepare a port so that it can be included in FreeBSD's ports
> collection. Anyway, I insist. I have some other more urgent tasks
> pending for the next month or so.
>
> >
> > I believe you are using the gnu compiler. What version?
> > (cc -v should tell you.)
>
> gcc version 2.7.2.1
>

Jon,

Just to verify things,
I built this on BSDI 3.0 with gcc version 2.7.2.1 without any problems.
This is on a Pentium machine of some sort. It is the closest thing I
have to your system, and frankly, I am surprised that you are having
problems. Same compiler, same cpu, pretty much the same OS, different
results? (BSDI is a Berkeley UNIX with support which has pretty much the
same roots as FreeBSD. Netcdf has been built on BSDI 2.x, which was
still binary compatible with FreeBSD?)

The only thing I can think of is gcc built improperly, or
that some include file values are hosed up, specifically the MAX_
stuff from limits.h or float.h, or DBL_EPSILON from float.h.
Another possibility is that configure is figure that you have signed
char when you don't, or vis versa.

Sorry I can't be of more help.

As I said in my previous message, I wouldn't worry too much about
failing nc_test and or nf_test in this case. The library is probably
fine.

-glenn