AC_CHECK_HEADER/AC_CHECK_DECL 2.69->2.69c probably changed quoting, breaks gimp as "...: ${+y}: bad substitution"

[Sergei Trofimovich]

The failure initially observed on gimp-2.10.22 source and fails
the same way on gimp from git.

Smaller example extracted from
    https://gitlab.gnome.org/GNOME/gimp/-/blob/master/configure.ac#L2131

$ cat configure.ac
  AC_INIT([GIMP], [gimp_version], 
[https://gitlab.gnome.org/GNOME/gimp/issues/new], [gimp])
  AC_PROG_CC

  dnl Taken from gimp as is:

  #######################
  # Check for Linux Input
  #######################

  AC_ARG_WITH(linux-input, [  --without-linux-input   don't build linux input 
event controller module])

  have_linux_input="no (linux input support disabled)"
  if test "x$with_linux_input" != "xno"; then
    AC_CHECK_HEADER(linux/input.h,
          AC_CHECK_DECL(KEY_OK,
                  have_linux_input=yes,
                  have_linux_input="no (needs Linux 2.6)",
                  [#include <linux/input.h>]))
  fi

  AC_OUTPUT

Reproducing:

  $ autoconf-2.69 && ./configure --host=x86_64-pc-linux-gnu
  ...
  configure: creating ./config.status
  <ok>

  $ autoconf-2.70_beta2 && ./configure --host=x86_64-pc-linux-gnu
  ./configure: line 1430: 5: Bad file descriptor
  checking whether  is declared... ./configure: line 1432: ${+y}: bad 
substitution
  <fails>

"bad substitution" looks like an autoconf macro expansion
problem. Warning: I don't understand m4.

gimp's use of macros looks underquoted but I would expect
behavior to be roughly the same between 2.69 and 2.69c.
Does it sound about right? 

Thank you!

-- 

  Sergei