[Python-modules-commits] r30280 - in packages/scipy/trunk/debian (4 files)

jtaylor-guest at users.alioth.debian.org jtaylor-guest at users.alioth.debian.org
Wed Aug 27 14:33:00 UTC 2014


    Date: Wednesday, August 27, 2014 @ 14:32:59
  Author: jtaylor-guest
Revision: 30280

add patches to fix test failures on i386 and mips:

  - fix-undefined-behavior-in-alngam.patch (Closes: #756905)
  - relax-bounds-of-interpolate-test.patch

Added:
  packages/scipy/trunk/debian/patches/fix-undefined-behavior-in-alngam.patch
  packages/scipy/trunk/debian/patches/relax-bounds-of-interpolate-test.patch
Modified:
  packages/scipy/trunk/debian/changelog
  packages/scipy/trunk/debian/patches/series

Modified: packages/scipy/trunk/debian/changelog
===================================================================
--- packages/scipy/trunk/debian/changelog	2014-08-27 14:18:43 UTC (rev 30279)
+++ packages/scipy/trunk/debian/changelog	2014-08-27 14:32:59 UTC (rev 30280)
@@ -1,3 +1,11 @@
+python-scipy (0.14.0-2) unstable; urgency=medium
+
+  * add patches to fix test failures on i386 and mips:
+    - fix-undefined-behavior-in-alngam.patch (Closes: #756905)
+    - relax-bounds-of-interpolate-test.patch
+
+ -- Julian Taylor <jtaylor.debian at googlemail.com>  Wed, 27 Aug 2014 14:40:24 +0200
+
 python-scipy (0.14.0-1) unstable; urgency=low
 
   [ Denis Laxalde ]

Added: packages/scipy/trunk/debian/patches/fix-undefined-behavior-in-alngam.patch
===================================================================
--- packages/scipy/trunk/debian/patches/fix-undefined-behavior-in-alngam.patch	                        (rev 0)
+++ packages/scipy/trunk/debian/patches/fix-undefined-behavior-in-alngam.patch	2014-08-27 14:32:59 UTC (rev 30280)
@@ -0,0 +1,21 @@
+Description: fix undefined behavior in alngam
+ INT(A) is undefined if A cannot be represented as INT
+ In MIPS the result is positive triggering a huge loop for:
+
+    scipy.stats.skellam.sf(np.inf, 10, 11)
+Date: Sun, 24 Aug 2014 15:10:56 +0200
+Author: Julian Taylor <jtaylor.debian at googlemail.com>
+Origin: 26f8a09d2ed6d123a8d624dffa5d3a5caf3d351e
+Bug: https://github.com/scipy/scipy/pull/3899
+Bug-Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=756905
+
+--- a/scipy/special/cdflib/alngam.f
++++ b/scipy/special/cdflib/alngam.f
+@@ -105,6 +105,7 @@ C
+ C     IF NECESSARY MAKE X AT LEAST 12 AND CARRY CORRECTION IN OFFSET
+ C
+ C
++      IF ((x.GT.12.0D0)) GO TO 90
+       n = int(12.0D0-x)
+       IF (.NOT. (n.GT.0)) GO TO 90
+       prod = 1.0D0

Added: packages/scipy/trunk/debian/patches/relax-bounds-of-interpolate-test.patch
===================================================================
--- packages/scipy/trunk/debian/patches/relax-bounds-of-interpolate-test.patch	                        (rev 0)
+++ packages/scipy/trunk/debian/patches/relax-bounds-of-interpolate-test.patch	2014-08-27 14:32:59 UTC (rev 30280)
@@ -0,0 +1,18 @@
+Description: relax bounds of interpolate test
+ A small rounding error appears with gcc-4.9 on i586 which causes the a
+ root to be -1e-19 instead of 0. and subsequently being cut by the exact
+ masking causing a testfailure.
+Bug: https://github.com/scipy/scipy/pull/3907
+Author: Julian Taylor <jtaylor.debian at googlemail.com>
+
+--- a/scipy/interpolate/tests/test_interpolate.py
++++ b/scipy/interpolate/tests/test_interpolate.py
+@@ -745,7 +745,7 @@ class TestPPoly(TestCase):
+         pp = PPoly.from_spline(spl)
+ 
+         r = pp.roots()
+-        r = r[(r >= 0) & (r <= 1)]
++        r = r[(r >= 0 - 1e-15) & (r <= 1 + 1e-15)]
+         assert_allclose(r, sproot(spl), atol=1e-15)
+ 
+     def test_roots_idzero(self):

Modified: packages/scipy/trunk/debian/patches/series
===================================================================
--- packages/scipy/trunk/debian/patches/series	2014-08-27 14:18:43 UTC (rev 30279)
+++ packages/scipy/trunk/debian/patches/series	2014-08-27 14:32:59 UTC (rev 30280)
@@ -3,3 +3,5 @@
 qhull-lfs.patch
 cap-ld-precision.patch
 fitpack-alias.patch
+fix-undefined-behavior-in-alngam.patch
+relax-bounds-of-interpolate-test.patch




More information about the Python-modules-commits mailing list