The reason I ask is I am trying to figure out
why you do not have xdr installed; is there a
package for mingw?
Sisyphus wrote:
----- Original Message ----- From: "Ed Hartnett"
(To Dennis: I'm not using Cygwin. I use the msys shell and the mingw
compiler to produce a native Windows library. I could probably achieve
the same on Cygwin, cross-compiling for native Win32. But my experience
has always been that a problem that arises using msys, will also arise
on a Cygwin cross-compile - unless, perhaps, that problem relates to
some tool or other that needs updating on *one* of the systems, but not
the other.)
Try configure with --disable-dap.
Does that also avoid the need for libcurl ?
Using --disable-dap does allow the configure step to complete successfully.
Then it's just a matter of hacking one's way through the compilation.
Building with my 64-bit mingw (gcc-4.7.0) compiler, first troublespot is
libdispatch/v2i.c (which used to be in libsrc). That file posed troubles
on 4.0.1, too. (Haven't struck any trouble with it on the 32-bit builds,
however.)
For 4.0.1 I inserted the following into v2i.c, immediately after the
first block of #includes:
#include "nc.h"
#include "rename.h"
#include <string.h>
#include <assert.h>
#include "rnd.h"
#include "utf8proc.h"
With 4.1.2, rename.h seems to have vanished, so I didn't include it -
but still had to insert the others. I'm no longer sure if *all* of those
files need to be included, but without that hack the NC struct doesn't
get defined.
Then:
Copy libsrc/onstack.h to libdispatch/onstack.h
Copy libsrc/rnd.h to libdispatch/rnd.h
Otherwise they don't get found when needed.
Then in libsrc/putget.c and libsrc/ncx.c:
typedef unsigned int uint;
As 'uint' is an unknown type. I guess it's an 'unsigned int'.
That's not too bad. Have I lost anything significant by '--disable-dap' ?
With 'make check -k' there are a few failures - see attached
'test.txt.gz' . The nctest.exe crashed - it was the only test that did
that. If all PDL::NetCDF tests pass, that's generally been good enough
for me. (I haven't yet verified that happens with this build of 4.1.2.)
For some comparison, I've attached the 4.0.1 test output (built using
the same compiler) - in test401.txt.gz. No crashes with that test suite.
If you have any questons/observations, I'll attend to them as time
permits - but I'm probably going to run out of time to do much more
until about Wednesday next week.
Thanks Ed !!
I tried a 32-bit build of 4.1.2, but configure wouldn't proceed when it
found out I don't have gfortran (though I do have g77). Does it really
need an f90 compiler ? If so, I can switch to a gcc-4.x.x compiler and
give that a try - but, for the moment at least, I'd like to stick with
gcc-3.4.5 for the 32-bit build if that's possible.
Cheers,
Rob
------------------------------------------------------------------------
_______________________________________________
netcdf-porting mailing list
netcdf-porting@xxxxxxxxxxxxxxxx
For list information, to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
