On Oct 8, 2010, at 6:41 PM, Eric Blake wrote: > AC_DEFUN([OMPI_F77_GET_ALIGNMENT, [ > AS_VAR_PUSHDEF([type_var], > m4_translit([[ompi_cv_f77_alignment_$1]], [*], [p])) > AS_VAR_SET([$2], [type_var])
Ah -- perfect; thanks! I found a few more cases of this issue (note that the AS_VAR_GET issue that started this thread has now been fixed; it's in the released Open MPI v1.4.3 and 1.5). Here's what I found -- I'll push these to OMPI SVN trunk tonight and relevant variants to v1.4/v1.5 branches: Index: f77_check_type.m4 =================================================================== --- f77_check_type.m4 (revision 23887) +++ f77_check_type.m4 (working copy) @@ -21,7 +21,10 @@ # OMPI_F77_CHECK_TYPE([type, action if found, action if not found]) # ----------------------------------------------------------------- AC_DEFUN([OMPI_F77_CHECK_TYPE],[ - AS_VAR_PUSHDEF([type_var], [ompi_cv_f77_have_$1]) + # Use of m4_translit suggested by Eric Blake: + # http://lists.gnu.org/archive/html/bug-autoconf/2010-10/msg00016.html + AS_VAR_PUSHDEF([type_var], + m4_translit([[ompi_cv_f77_have_$1]], [*], [p])) # Determine Fortran datatype size. # First arg is type, 2nd arg is config var to define Index: f90_get_sizeof.m4 =================================================================== --- f90_get_sizeof.m4 (revision 23887) +++ f90_get_sizeof.m4 (working copy) @@ -21,7 +21,10 @@ # OMPI_F90_GET_SIZEOF(type, variable to set) # ------------------------------------------ AC_DEFUN([OMPI_F90_GET_SIZEOF],[ - AS_VAR_PUSHDEF([type_var], [ompi_cv_f90_sizeof_$1]) + # Use of m4_translit suggested by Eric Blake: + # http://lists.gnu.org/archive/html/bug-autoconf/2010-10/msg00016.html + AS_VAR_PUSHDEF([type_var], + m4_translit([[ompi_cv_f90_sizeof_$1]], [*], [p])) AC_CACHE_CHECK([size of Fortran 90 $1], type_var, [OMPI_F77_MAKE_C_FUNCTION([ompi_ac_size_fn], [size]) Index: f77_get_sizeof.m4 =================================================================== --- f77_get_sizeof.m4 (revision 23887) +++ f77_get_sizeof.m4 (working copy) @@ -21,7 +21,10 @@ # OMPI_F77_GET_SIZEOF(type, variable to set) # ------------------------------------------ AC_DEFUN([OMPI_F77_GET_SIZEOF],[ - AS_VAR_PUSHDEF([type_var], [ompi_cv_f77_sizeof_$1]) + # Use of m4_translit suggested by Eric Blake: + # http://lists.gnu.org/archive/html/bug-autoconf/2010-10/msg00016.html + AS_VAR_PUSHDEF([type_var], + m4_translit([[ompi_cv_f77_sizeof_$1]], [*], [p])) AC_CACHE_CHECK([size of Fortran 77 $1], type_var, [OMPI_F77_MAKE_C_FUNCTION([ompi_ac_size_fn], [size]) Index: f77_get_alignment.m4 =================================================================== --- f77_get_alignment.m4 (revision 23887) +++ f77_get_alignment.m4 (working copy) @@ -23,7 +23,10 @@ AC_DEFUN([OMPI_F77_GET_ALIGNMENT],[ unset happy OMPI_VAR_SCOPE_PUSH([happy ompi_conftest_h]) - AS_VAR_PUSHDEF([type_var], [ompi_cv_f77_alignment_$1]) + # Use of m4_translit suggested by Eric Blake: + # http://lists.gnu.org/archive/html/bug-autoconf/2010-10/msg00016.html + AS_VAR_PUSHDEF([type_var], + m4_translit([[ompi_cv_f77_alignment_$1]], [*], [p])) AC_CACHE_CHECK([alignment of Fortran $1], type_var, [OMPI_F77_MAKE_C_FUNCTION([ompi_ac_align_fn], [align]) Index: f90_check_type.m4 =================================================================== --- f90_check_type.m4 (revision 23887) +++ f90_check_type.m4 (working copy) @@ -20,7 +20,10 @@ # OMPI_F90_CHECK_TYPE([type, action if found, action if not found]) # ----------------------------------------------------------------- AC_DEFUN([OMPI_F90_CHECK_TYPE],[ - AS_VAR_PUSHDEF([type_var], [ompi_cv_f90_have_$1]) + # Use of m4_translit suggested by Eric Blake: + # http://lists.gnu.org/archive/html/bug-autoconf/2010-10/msg00016.html + AS_VAR_PUSHDEF([type_var], + m4_translit([[ompi_cv_f90_have_$1]], [*], [p])) # Determine Fortran datatype size. # First arg is type, 2nd arg is config var to define -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
