Compiler preprocessor defines

GNU Radio
1

Is there any preprocessor define (or defines) that can be used
when compiling C++ GnuRadio code to differentiate between
different OSs? I’ve looked through config.h but didn’t see
anything obvious. I need to compile two different pieces
of code for OSX vs LINUX.

@(^.^)@ Ed

2

On Mon, Nov 03, 2008 at 02:16:10PM -0500, Ed Criscuolo wrote:

Is there any preprocessor define (or defines) that can be used
when compiling C++ GnuRadio code to differentiate between
different OSs? I’ve looked through config.h but didn’t see
anything obvious. I need to compile two different pieces
of code for OSX vs LINUX.

FWIW, the “autoconf way” is to test for features, not OS’s.
In any event, if you look through config/*.m4 directory you’ll find
all sorts of examples.

See also the autoconf manual.

What are you trying to conditionalize?

Eric

3

Eric B. wrote:

all sorts of examples.

See also the autoconf manual.

What are you trying to conditionalize?

The tun/tap pseudo device is implemented very differently
on OSX vs LINUX, UNIX, et al.

Under LINUX et al, you open a single character device, usually
located at /dev/net/tun. Then, you use ioctl’s to to select
the pseudo network device name (tun0, tun1, tap0, tap1, etc) and
set the type (tun vs tap). There’s also an option for requesting
a name with the next available number (ie - tun%d or tap%d).

Under OSX, the tun/tap driver precreates the character devices
/dev/tun0 thru /dev/tun15 and /dev/tap0 thru /dev/tap15. The
application opens the appropriate character device to create the
associated pseudo network device of the same name and type.

@(^.^)@ Ed

4

Discuss-gnuradio mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/discuss-gnuradio

5

On Mon, Nov 03, 2008 at 03:46:56PM -0500, Ed Criscuolo wrote:

In any event, if you look through config/*.m4 directory you’ll find
all sorts of examples.

See also the autoconf manual.

What are you trying to conditionalize?

The tun/tap pseudo device is implemented very differently
on OSX vs LINUX, UNIX, et al.

OK.

Take a look at how we handle the “Fast USB” technique selection.
See config/usrp_fusb_tech.m4.

It would probably be a good idea to abstract the whole “open the
tap/tun device” operation into a separate function, then swig it up so
that it’s accessible from python.

Eric

6

The general way to do this is to create an m4 file with a ‘case’ in it
based on the “$host_os”, e.g. (as taken and expanded from
grc_gr_audio_osx.m4; the case strings may or not be entirely correct,
and are used as examples):

passed=yes
[set some other global variables to expected values]
case “$host_os” in
darwin*)
[set some variable here for Darwin only]
;;
freebsd*)
[set some variable here for freebsd only]
;;
linux*)
[set some variable here for linux only]
;;
*)
[default, maybe print something?; set some variable?]
passed=no
;;
esac
[perform checks on the variables, to make sure they work together; set
“passed=no” on failure]
if test $passed = yes; then
AC_MSG_RESULT ([passed; using tap/tun FOO])
else
AC_MSG_RESULT ([failed; not using tap/tun])
fi

7

Michael D. wrote:

;;
esac
OSX implements tap/tun between 10.4 and 10.5 … in that case, there
would be different ‘darwin8*’ and ‘darwin9*’ cases. The same might hold
between Linux kernels 2.4.X and 2.6.X, and I’m sure there’s a way to
create cases for each of those.

I’ve opened ticket #311 to capture this enhancement request, and
associated suggestions.

@(^.^)@ Ed