Test compile: mpn_lshift_com optimization 2

[Richard Biener]

On Wed, 4 Jan 2017, Niels Möller wrote:

> Richard Biener <rguenther at suse.de> writes:
> 
> > This test intoduced with GMP 6.1.2 FAILs with malloc debugging
> > because it appearantly relies on malloc returing zeroed memory?
> > Or it tests the wrong array elements?
> 
> Can you please clarify which test program fails? There's no "t.c" or
> "a.out" files in my gmp tree. Please also provide the additional
> information requested in the Bug reporting instructions in the manual.
> 
> For what it's worth, 
> 
>   TESTS_ENVIROMENT=valgrind make check
> 
> works fine in my gmp build.

Our build environment has

export MALLOC_CHECK_=3
export MALLOC_PERTURB_=69

and the patch below fixes the issue for me (after re-generating configure 
of course).  Otherwise configury fails to find a working compiler.

It fails on all archs (x86_64, i586, aarch64, s390x, ppc64le).

Thanks,
Richard.

--- acinclude.m4.orig   2017-01-04 11:40:34.672561902 +0100
+++ acinclude.m4        2017-01-04 11:40:42.324648573 +0100
@@ -741,7 +741,7 @@
   long i;
   for (i = 0; i < 88 + 1; i++)
     a[i] = ~0L;
-  r = malloc (10000 * sizeof (unsigned long));
+  r = calloc (10000, sizeof (unsigned long));
   r2 = r;
   for (i = 0; i < 528; i += 23)
     {