[PATCH] warnings.m4: provide a means to specify the program to compile

[Akim Demaille]

Hi!

I have submitted this patch several times in this thread (starting at
http://lists.gnu.org/archive/html/bug-gnulib/2012-04/msg00049.html),
and for some reason it remains unnoticed, and uncommented.  Please let
me know what is wrong in my approach.

I would like to wrap a Bison 2.5.1 soon, and I expect to use this
patch to decide whether I enable -Wzero-as-null-pointer-constant (as
G++ can happily reject 0 as a pointer without supporting nullptr, not
considered as a bug by G++ people).

The attached patch enhances warnings.m4 in several ways:

- gl_COMPILER_OPTION_IF allows to define more fine grain tests on the
  behavior of the compiler.  Forcing the result to be an assignment to
  a variable (which is AC_SUBST'ed) does not seem to offer the
  orthogonal design one would expect.  This would be useful elsewhere,
  for instance in manywarnings, as I have already reported
  (http://lists.gnu.org/archive/html/bug-gnulib/2012-03/msg00198.html)

- separating the diagnostic part from the action part provides a
  cleaner way to implement test cases.

- separating the diagnostic part from the action part helps extending
  both in a cleaner way:

  - it allowed me to cleanly integrate the missing AC_SUBST for
    WARN_CFLAGS instead of leaving this to modules/warnings

  - it will make it easier to check for the answers of the compiler on
    stderr (as currently, we merely check whether the compiler rejects
    the candidate option with an error status; this results in uselessly
    long compilation command lines, and spurious warnings in some
    situations).

- the added parameter to gl_WARN_ADD allows to fine tune whether a
  given warning is wanted or not (e.g., 0 vs. nullptr as above).

If there are flaws to address in this patch, I'd be happy to.

        Akim

0001-warnings.m4-provide-a-means-to-specify-the-program-t.patch
Description: Binary data