Re: macros which define macros

[Eric Blake]

Sam Steingold <sds <at> gnu.org> writes:

> 
> are you watching CLISP development? 

Nope.  Should I be? :)

> | m4_define([CL_DEFUN_CHECKER],
> |   [_$0([$]1, [$]2, [$]3, [$]4, [$]5, [$]6, [$1])])
> 
> interesting.
> the above version is shorter and easier for me (and, more importantly,
> those who will come after me) to understand, so I will stick with it.

Good - the first approach is certainly more idiomatic in autoconf.  I merely 
threw out the second approach for teaching purposes, based on the subject line; 
and I'm not at all heart-broken that you aren't using it (while writing my 
first reply, it took the most time and brain-effort to write the second 
approach, for the very reason that it is not very idiomatic).

> | cl_program_xcheck=`cat <<\EOF
> | AC_LANG_PROGRAM([#include <sys/socket.h>
> | #include <X11/Xlib.h>
> | #include <X11/Xauth.h>],
> | [XauGetAuthByAddr(AF_INET,0,"",13,"localhost:0.0",13,"localhost:0.0");])
> | EOF`

Bug on my part (and perhaps we need to update the manual) - not all shells 
behave the same when presented with incomplete here-docs:

POSIX states that this is undefined:

foo=`cat <<EOF
`

since the ` occurs unquoted within a here-document.

POSIX also requires that a here-doc end when the delimiter occurs on a line 
with a newline.  But various shells interpret this differently:

foo=`cat <<EOF
hi
EOF`

ash claims that EOF is not on a line by itself, thus the here-doc is not 
terminated.  bash claims that ` terminates the command substitution, and that a 
here-doc can end on end-of-input in addition to a newline.  I agree with ash's 
interpretation that this is not the end of the here-doc, but that means that 
behavior is undefined, which means bash's behavior is also acceptable.

On the other-hand, this should be well-defined by POSIX:

foo=$(cat <<EOF
hi
EOF)

EOF does not occur on a line by itself, so you could argue that a strict 
reading of POSIX says that particular ) should not terminate the command 
substitution because the contents of $() using that ) does not form a valid 
script.  But again, ash is the only shell that rejected that.

To be portable, you need:

foo=`cat <<EOF
hi
EOF
`

> 
> this expands to
> ~    cl_clx_XauGetAuthByAddr_program=`cat <<\EOF

Looks like inadvertant macro expansion in the variable name...

> /* confdefs.h.  */
> _ACEOF
> cat confdefs.h >>conftest.$ac_ext
> cat >>conftest.$ac_ext <<_ACEOF

Oh well.  It looks like stashing the program in a shell variable might not work 
(it certainly isn't as easy as I had thought it might be).  But using an m4 
macro to reduce duplication in your source should still be viable.

-- 
Eric Blake