Re: [gforth] Compiling with ffcall-1.9

[Dennis Ruffer]

How do I tell if I have libffi? I don't see anything in the usual places. DaR

On Mar 28, 2004, at 12:39 PM, Anton Ertl wrote:

Bernd Paysan wrote:

On Sunday 28 March 2004 01:33, Dennis Ruffer wrote:

In case anyone wants to investigate what I mentioned in com.lang.forth,
here's what I get when ffcall-1.9 is installed on Panther.

This is not a ffcall problem.

My darwin script looks like this:

[...]

checking for long long... yes
checking size of long long... (cached) 0

[...]

checking for a C type for double-cells... none
configure: WARNING: Emulating double-cell arithmetic. This may be slow.

[...]

In file included from engine.c:337:
./prim: In function `engine':
./prim:1983: warning: implicit declaration of function `ecvt'
./prim:1983: warning: assignment makes pointer from integer without a
cast
./prim:2309: error: aggregate value used where an integer was expected
./prim:2328: error: aggregate value used where an integer was expected
./prim:2360: error: incompatible types in assignment
./prim:2393: error: incompatible types in assignment
./prim:2417: error: incompatible types in assignment
./prim:2417: warning: value computed is not used
make[2]: *** [engine.o] Error 1
make[1]: *** [engines] Error 2
make: *** [kernel/version.fs] Error 2

The problem seems to be that "long long" (or sizeof) is broken, because
sizeof(long long) gives 0.

The reason for this configure result is that I explicitly programmed
it into configure.in, because long long is broken with at least some
gcc-2.95s on PPC.

The errors above will happen whenever the double type is not long
long, i.e. on all I32LP64 systems in addition to the PPCs.  Someone
should #ifdef the code of AV_LONGLONG etc. for this cases, just like
D+ has an #ifdef for this case.

BTW, libffi seems to be part of recent gccs and seems to be installed
when you install gcc.  Some of the thoughts that come to my mind:

- libffi may be better maintained and/or have more targets than ffcall
(not checked).

- libffi would often require less effort than ffcall from the gforth
installer.

- How much more effort is it to use libffi?

- How about using them alternatively (whichever is available on
installation)?  Do both fit with our higher-level interface?

- anton

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

smime.p7s
Description: S/MIME cryptographic signature