Cleanup of memory allocation in the tests:

  * only tests_allocate, tests_reallocate and tests_free should be used
    (now exported instead of mpfr_default_allocate, etc.);
  * this is now mentioned in doc/README.dev;
  * added checks to tools/mpfrlint (heuristics as usual).

Note: This follows the discussion at
  https://sympa.inria.fr/sympa/arc/mpfr/2015-12/msg00055.html
and should solve these problems under MS Windows (mingw32), though this
cleanup is more general than an attempt to fix a Windows issue.

git-svn-id: svn://scm.gforge.inria.fr/svn/mpfr/trunk@9891 280ebfd0-de03-0410-8827-d642c229c3f4

1 files changed, 2 insertions, 7 deletions

diff --git a/tests/tgrandom.c b/tests/tgrandom.c
index c8f452cb0..75dccc0c9 100644
--- a/tests/tgrandom.c
+++ b/tests/tgrandom.c
@@ -56,12 +56,7 @@ test_grandom (long nbtests, mpfr_prec_t prec, mpfr_rnd_t rnd,
   int i, inexact;
 
   nbtests = (nbtests & 1) ? (nbtests + 1) : nbtests;
-  t = (mpfr_t *) malloc (nbtests * sizeof (mpfr_t));
-  if (t == NULL)
-    {
-      fprintf (stderr, "tgrandom: can't allocate memory in test_grandom\n");
-      exit (1);
-    }
+  t = (mpfr_t *) tests_allocate (nbtests * sizeof (mpfr_t));
 
   for (i = 0; i < nbtests; ++i)
     mpfr_init2 (t[i], prec);
@@ -106,7 +101,7 @@ test_grandom (long nbtests, mpfr_prec_t prec, mpfr_rnd_t rnd,
 
   for (i = 0; i < nbtests; ++i)
     mpfr_clear (t[i]);
-  free (t);
+  tests_free (t, nbtests * sizeof (mpfr_t));
   return;
 }