[libstdc++] [testsuite] require cmath for [PR114359]

When !_GLIBCXX_USE_C99_MATH_TR1, binomial_distribution doesn't use the
optimized algorithm that was fixed in response to PR114359.  Without
that optimized algorithm, operator() ends up looping very very long
for the test, to the point that it would time out by several orders of
magnitude, without even exercising the optimized algorithm that we're
testing for regressions.  Arrange for the test to be skipped if that
bit won't be exercised.


for  libstdc++-v3/ChangeLog

	PR libstdc++/114359
	* testsuite/26_numerics/random/binomial_distribution/114359.cc:
	Require cmath.

1 files changed, 13 insertions, 0 deletions

diff --git a/libstdc++-v3/testsuite/26_numerics/random/binomial_distribution/114359.cc b/libstdc++-v3/testsuite/26_numerics/random/binomial_distribution/114359.cc
index c1e4c38..12d967d 100644
--- a/libstdc++-v3/testsuite/26_numerics/random/binomial_distribution/114359.cc
+++ b/libstdc++-v3/testsuite/26_numerics/random/binomial_distribution/114359.cc
@@ -2,6 +2,19 @@
 
 // Bug 114359 - std::binomial_distribution hangs in infinite loop
 
+// { dg-require-cmath "" }
+
+// The requirement above is not strictly true.  The test should work
+// without cmath, and it probably does, but without cmath,
+// binomial_distribution::operator() skips the optimized algorithm and
+// calls _M_waiting to loop a gazillion times.  On aarch64-rtems6
+// qemu, that loop takes over 5 minutes to go through a small fraction
+// of the iteration space (__x at 22k, limited at 1G; __sum at 2e-5,
+// limited at 0.69).  The bug we're regression-testing here was in the
+// cmath-requiring bit, so even if this could conceivably not time out
+// on a really fast machine, there's hardly any reason to exercise
+// this extreme case.
+
 #include <random>
 
 int main()