Re: [PATCH] {maint} Document in detail some limitations of aclocal.

[Eric Blake]

On 11/04/2010 03:28 PM, Stefano Lattarini wrote:
>>> +# Indirection used here, to avoid triggering the bug described
>>> +# above.
>>> +AC_DEFUN([MY_DEFUN], [m4_apply([AC_DEFUN], [$1], [$2])])
>>
>> This is insufficiently quoted, it should be:
>>                         [m4_apply([AC_DEFUN], [[$1], [$2]])]
> I'm not an m4 quoting guru, but I think that one single level of quotes
> is sufficient to avoid *spurious* expansions.  Do you have any reason or
> example showing that two levels of quotes are indeed needed?

Yes - the documentation of m4_apply includes this example:

m4_apply([m4_count], [])
0
m4_apply([m4_count], [[]])
1
m4_apply([m4_count], [[1], [2]])
2
m4_apply([m4_join], [[|], [1], [2]])
1|2

That is, m4_apply is implemented as $1($2).  Notice that the entire
second argument to m4_apply is passed as a series of arguments to $1, so
each argument to $1 must be quoted, AND that series of arguments to
m4_apply must be quoted into a single unit.

The way you wrote it, you would have called AC_DEFUN($1), and $2 would
be a spurious argument (not sure if it would be silently ignored or
trigger a warning), when you really wanted to call AC_DEFUN([$1], [$2]).

-- 
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature