Kimball Thurston <[EMAIL PROTECTED]> writes:
> I added a more verbose fatal message, and the class that we are looking
> for has an empty name (class->name is empty...) I'm not sure what that
> means....
Do you mean a NULL-pointer, or a pointer to an empty string? In any
case, it is not normal. Classes are automatically generated (in the .x
files), and usually look something like (from pad.c.x):
struct lsh_class packet_pad_class =
{ STATIC_HEADER,
/* ^object header */
&abstract_write_pipe_class, "packet_pad", sizeof(struct packet_pad),
/* ^superclass ^ name field ^size */
do_packet_pad_mark,
/* ^mark function */
NULL
/* ^free function, if needed */
};
It seems really strange that the *class* object is invalid. The
classes passed to the typechecking functions are addresses of known
statically allocated objects. In gaba_apply(), (which was the failing
function, iirc), it passes &command_simple_class, which should be
defined in command.h.x. Perhaps I should hack the generator to make
all classes as const, just in case?
> If you look in gmp.h, it has a ifdef block for extern inline - maybe you
> could borrow from that, but like I said, thing seemed to be happy, so
> I'm not going to worry about it... it may have been from before where
> inline wasn't getting properly defined or something...
Are you saying that AC_C_INLINE fix this problem as well? Then I'll
forget about it for now. Let me know if it appears agin.
> Oh, another side note, you may want to mention that you need guile and
> scsh and m4 for things to work. The first time I downloaded stuff, it
> took me a bit to figure out that guile and scsh were needed... (when the
> configure was puking...)
You'll need either scsh, or guile+slib. And an m4 that can deal with
eight-bit quote characters (such as ��). But you shouldn't need any of
that for a basic "./configure && make" install, only if you end up
hacking any of the source files.
/Niels
