Cleanup / fixes related to intmax_t, mpfr_intmax_t, etc.

The issues were:
  * The build of the tests (make check) could fail with CFLAGS implying
    "-std=c90 -Werror=format", i.e. if intmax_t was available but the
    use of the "j" printf length modifier made the compilation fail
    (its unconditional use was incorrect anyway).
  * MPFR_INTMAX_MAX had definitions in both mpfr-impl.h & mpfr-intmax.h,
    with inconsistent usage.

Detailed changes:
  * acinclude.m4: removed the definition of MPFR_PRINTF_MAXLM, which
    could be incorrect (large numbers were not tested to select its
    value); in the MPFR_FUNC_GMP_PRINTF_SPEC macro, check sprintf()
    in addition to gmp_sprintf() as it is now also used in the tests;
    updated some comments and messages.
  * src/mpfr-impl.h: removed the MPFR_*INTMAX* definitions; they had
    been added due to
      https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=582698
    with -ansi -pedantic-errors, but this issue was fixed in GCC 4.8
    (released 6 years ago), so that they are no longer needed.
  * src/mpfr-intmax.h: completed the definitions; mpfr_intmax_t is now
    the largest integer type available and fully working (in particular
    with printf), thus it may be smaller than intmax_t.
  * doc/README.dev: updated the description of mpfr_intmax_t, etc.
  * tests/tversion.c: output sizeof(mpfr_intmax_t) too.
  * Various .c files in src and tests related to (u)intmax_t: use
    INTMAX_MAX instead of MPFR_INTMAX_MAX, etc. as Debian bug 582698
    (mentioned above) no longer occurs and mpfr_intmax_t may be smaller
    than intmax_t; use the "j" printf length modifier instead of
    MPFR_PRINTF_MAXLM, and protect it with "#ifndef NPRINTF_J"
    (only the tests are concerned, so that this is not an issue).

git-svn-id: svn://scm.gforge.inria.fr/svn/mpfr/trunk@13457 280ebfd0-de03-0410-8827-d642c229c3f4

1 files changed, 12 insertions, 15 deletions

diff --git a/doc/README.dev b/doc/README.dev
index 85c8c9b2c..9fe4d4173 100644
--- a/doc/README.dev
+++ b/doc/README.dev
@@ -1168,21 +1168,18 @@ array of bits (e.g. for limbs) and to temporarily avoid integer overflow.
 
 ===========================================================================
 
-To use the features related to intmax_t, "mpfr-intmax.h" must be included
-before "mpfr-impl.h", to express the need for the largest integer types.
-Then, these features can be used only if _MPFR_H_HAVE_INTMAX_T is defined.
-In such a case, do not use the macros UINTMAX_MAX, INTMAX_MAX and INTMAX_MIN
-directly (because they might make the compilation fail due to bugs in the
-build environment), but the MPFR versions: MPFR_UINTMAX_MAX, MPFR_INTMAX_MAX
-and MPFR_INTMAX_MIN. Whether _MPFR_H_HAVE_INTMAX_T is defined or not, these
-macros and the types mpfr_intmax_t / mpfr_uintmax_t are always defined, and
-they correspond to the latest integer types available (e.g. intmax_t, else
-long long, else long).
-
-It can be assumed that any signed (resp. unsigned) integer type can be
-represented in mpfr_intmax_t (resp. mpfr_uintmax_t).
-
-The corresponding length specifier is MPFR_INTMAX_FSPEC.
+To use features related to types larger than type long, "mpfr-intmax.h"
+must be included before "mpfr-impl.h".
+
+The intmax_t and uintmax_t types can be used only if _MPFR_H_HAVE_INTMAX_T
+is defined. In this case, the printf / gmp_printf length specifier "j" can
+be used only when NPRINTF_J is not defined.
+
+For internal use, mpfr-intmax.h also unconditionally defines mpfr_intmax_t,
+mpfr_uintmax_t, MPFR_UINTMAX_MAX, MPFR_INTMAX_MAX, MPFR_INTMAX_MIN and
+the corresponding length specifier MPFR_INTMAX_FSPEC.
+
+Warning! mpfr_intmax_t may be smaller than intmax_t if NPRINTF_J is defined.
 
 ===========================================================================