This is the mail archive of the mailing list for the glibc project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Disable one hypot test for inline double testing

On 05/05/2012 05:57 PM, Joseph S. Myers wrote:
I saw one test of hypot fail in test-idouble on x86 because of
overflow to infinity.  I don't know what perturbed the inline code
generation for this test, but it seems reasonable to disable it for
inline double testing the same way other similar tests are disabled
for inline float and long double testing; I propose this patch (tested
x86) to do so.

GCC might make different inline decisions based on the size of functions. That's why the lrint failure I noticed with GCC 4.7 came in - we now have too many tests and therefore GCC makes different choices for inlining some of the tests into main (they are called once).

This patch is fine,

2012-05-05 Joseph Myers<>

	* math/ (hypot_test) [TEST_DOUBLE&&  TEST_INLINE]:
	Disable one test.

diff --git a/math/ b/math/
index 0875e2c..ceb7d35 100644
--- a/math/
+++ b/math/
@@ -4049,7 +4049,7 @@ hypot_test (void)
    TEST_ff_f (hypot, 0x1.234566p-126L, 0x1.234566p-126L, 1.891441686191081936598531534017449451173e-38L);

-#ifndef TEST_FLOAT
+#if !defined TEST_FLOAT&&  !(defined TEST_DOUBLE&&  defined TEST_INLINE)
    TEST_ff_f (hypot, 0x3p1021L, 0x4p1021L, 0x5p1021L);

 Andreas Jaeger aj@{,} Twitter/Identica: jaegerandi
  SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GF: Jeff Hawn,Jennifer Guild,Felix Imendörffer,HRB16746 (AG Nürnberg)
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]