Imath
Imath copied to clipboard
AcademySoftwareFoundation
Test failure on Fedora i686
Interestingly the arm 32 bit build passes and all 64 bit platforms pass, but on i686 I get the following:
24/36 Test #20: Imath.testFun ...........................Subprocess aborted***Exception: 0.09 sec
Testing functions in ImathFun.h
floor
ceil
trunc
divs / mods
divp / modp
successor, predecessor
f 0
sf 1.40129846e-45
pf -1.40129846e-45
spf -0
psf 0
f -0
sf 1.40129846e-45
pf -1.40129846e-45
spf -0
psf 0
f 1
sf 1.00000012
pf 0.99999994
spf 1
psf 1
f -1
sf -0.99999994
pf -1.00000012
spf -1
psf -1
f 16
sf 16.0000019
pf 15.999999
spf 16
psf 16
f 7
sf 7.00000048
pf 6.99999952
spf 7
psf 7
f 0.699999988
sf 0.700000048
pf 0.699999928
spf 0.699999988
psf 0.699999988
f inf
sf inf
pf inf
spf inf
psf inf
f -inf
sf -inf
pf -inf
spf -inf
psf -inf
f nan
sf nan
pf nan
spf nan
psf nan
ImathTest: /builddir/build/BUILD/Imath-3.0.2/src/ImathTest/testFun.cpp:70: void testf(float, bool): Assertion `bit_cast<unsigned> (pf) == bit_cast<unsigned> (f)' failed.
Full log here: https://kojipkgs.fedoraproject.org//work/tasks/9350/68939350/build.log
32 bits test are failing here too with for example:
ImathTest: /home/elrick/Temp/Imath/src/ImathTest/testInterval.cpp:189: void {anonymous}::testExtendByPoint(const char*) [with T = float]: Assertion `b.min == min && b.max == max' failed.
Is 32 bits i686 supported at all? I'm just a packager and I'd like to know if it's worth building the 32 bits version.
Well for Fedora there are discussions about limiting 32bit support. Currently there is no installer for 32bit but some libraries are maintained for programs that are not available for 64bit, like Steam. While that particular one doesn't apply here it's hard to make assumptions on what 3rd party software end users need.
It appears the NaN isn't being detected correctly by predf(). Maybe it would be safer to use isinff(f) || isnanf(f) here and here. I wonder if this is related to AcademySoftwareFoundation/openexr#346. Compiling with -ffloat-store may also fix this.
That's a head-scratcher. I just pushed another commit to #209 with some extra debugging info. @hobbes1069, could you try it again, and post the contents of LastTest.log? Thanks.
Trying now. I was surprised as well, I triple checked that the patch was being applied in the spec file.
Hopefully this helps but it doesn't look any different.
https://hobbes1069.fedorapeople.org/LastTest.log
That does provide some clues, although it is quite strange. Both predf(0x7f800001) and succf(0x7f800001) return 0x7fc00001.
That's a great clue Cary. The successor and predecessor of NAN is NAN. 0x7f800001 is a signaling NAN, and 0x7fc00001 is a Quiet NAN. (c.f. https://developer.arm.com/documentation/dui0475/g/BABFGDBH) An implementation is allowed to convert a signaling nan to a quiet nan, if signaling nans are not supported, and that seems to be what's happening here. (http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1011.htm). The standards committee recommended, in that document, not supporting signaling nans in language implementations, and also recommended a macro to indicate if they are supported.
OK, this is almost certainly a mistaken assumption in the test, assuming that the bits don't change when in fact, the system may convert the signalling NaN to a quiet NaN by simply copying the value. I pushed another commit that changes the test to simply confirm that if isnan(f), then isnan(sf) and isnan(pf), which seems to validate the behavior even if the bits are different.
@hobbes1069, could you try it again? And thanks for your patience in doing the testing.
So we need unit tests for the unit tests? :)
Still bombed, I just pushed the file to the same link above.
Gah, it made it past the float tests but failed on the doubles because I made a copy/paste typo between testf() and testd(). Can you try it once more?
I put some further debugging info in there to better understand what's happening, could you repost the LastTest.log? Thanks again.
I can confirm that the following tests are failling on i686 (void-linux here):
The following tests FAILED:
1 - PyImath.PyImathTest_Python3 (Failed)
18 - Imath.testMatrix (Subprocess aborted)
23 - Imath.testInterval (Subprocess aborted)
32 - Imath.testBoxAlgo (Subprocess aborted)
33 - Imath.testBox (Subprocess aborted)
Test1:
Running testV2Array
V2iArray
ok
V2fArray
Traceback (most recent call last):
File "/builddir/Imath-3.1.5/src/python/PyImathTest/pyImathTest.in", line 10137, in <module>
test[1]()
File "/builddir/Imath-3.1.5/src/python/PyImathTest/pyImathTest.in", line 2619, in testV2Array
testV2xArray (V2fArray, V2f, FloatArray)
File "/builddir/Imath-3.1.5/src/python/PyImathTest/pyImathTest.in", line 2588, in testV2xArray
assert (l[0] == v0.length())
AssertionError
18/38 Test #18: Imath.testMatrix ........................Subprocess aborted***Exception: 0.00 sec
Testing functions in ImathMatrix.h
Imath::M22f constructors and equality operators
M22d constructors and equality operators
M22f inversion operators
ImathTest: /builddir/Imath-3.1.5/src/ImathTest/testMatrix.cpp:116: void testMatrix(): Assertion `m3 == m2' failed.
32/38 Test #32: Imath.testBoxAlgo .......................Subprocess aborted***Exception: 2.35 sec
Testing box algorithms
ray-box entry and exit, random rays
box = ((-1 -1 -1) (1 1 1))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((10 20 30) (1010 21 31))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((10 20 30) (11 1020 31))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((10 20 30) (11 21 1030))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((-1e+10 -2e+10 -3e+10) (5e+15 6e+15 7e+15))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (2 1 1))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (1 2 1))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (1 1 2))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (1 2 3))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (2 3 1))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (2 1 3))
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((-1 -2 1) (-1 -2 1))
single-point box, ray intersects
single-point box, ray does not intersect
box = ((1 1 1) (1 1 1))
single-point box, ray intersects
single-point box, ray does not intersect
box = ((0 0 0) (0 0 0))
single-point box, ray intersects
single-point box, ray does not intersect
empty box, no rays intersect
ray-box entry and exit, nearly axis-parallel rays
dir ~ (1 0 0), result = 1
dir ~ (-1 0 0), result = 1
dir ~ (1 0 0), result = 0
dir ~ (-1 0 0), result = 0
dir ~ (0 1 0), result = 1
dir ~ (0 -1 0), result = 1
dir ~ (0 1 0), result = 0
dir ~ (0 -1 0), result = 0
dir ~ (0 0 1), result = 1
dir ~ (0 0 -1), result = 1
dir ~ (0 0 1), result = 0
dir ~ (0 0 -1), result = 0
ray-box intersection, random rays
box = ((-1 -1 -1) (1 1 1))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((10 20 30) (1010 21 31))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((10 20 30) (11 1020 31))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((10 20 30) (11 21 1030))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((-1e+10 -2e+10 -3e+10) (5e+15 6e+15 7e+15))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (2 1 1))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (1 2 1))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (1 1 2))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (1 2 3))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (2 3 1))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((1 1 1) (2 1 3))
ray starts inside box
ray starts outside box, intersects
ray starts outside box, does not intersect
box = ((-1 -2 1) (-1 -2 1))
single-point box, ray intersects
single-point box, ray does not intersect
box = ((1 1 1) (1 1 1))
single-point box, ray intersects
single-point box, ray does not intersect
box = ((0 0 0) (0 0 0))
single-point box, ray intersects
single-point box, ray does not intersect
empty box, no rays intersect
ray-box intersection, nearly axis-parallel rays
dir ~ (1 0 0), result = 1
dir ~ (-1 0 0), result = 1
dir ~ (1 0 0), result = 0
dir ~ (-1 0 0), result = 0
dir ~ (0 1 0), result = 1
dir ~ (0 -1 0), result = 1
dir ~ (0 1 0), result = 0
dir ~ (0 -1 0), result = 0
dir ~ (0 0 1), result = 1
dir ~ (0 0 -1), result = 1
dir ~ (0 0 1), result = 0
dir ~ (0 0 -1), result = 0
transform box by matrix
ImathTest: /builddir/Imath-3.1.5/src/ImathTest/testBoxAlgo.cpp:816: void {anonymous}::boxMatrixTransform(): Assertion `b21 == b2' failed.
Start 33: Imath.testBox
33/38 Test #33: Imath.testBox ...........................Subprocess aborted***Exception: 0.00 sec
Testing box methods
constructors for type V2s
constructors for type V2i
constructors for type V2i64
constructors for type V2f
constructors for type V2d
constructors for type V3s
constructors for type V3i
constructors for type V3i64
constructors for type V3f
constructors for type V3d
constructors for type V4s
constructors for type V4i
constructors for type V4i64
constructors for type V4f
constructors for type V4d
makeEmpty() for type V2s
makeEmpty() for type V2i
makeEmpty() for type V2i64
makeEmpty() for type V2f
makeEmpty() for type V2d
makeEmpty() for type V3s
makeEmpty() for type V3i
makeEmpty() for type V3i64
makeEmpty() for type V3f
makeEmpty() for type V3d
makeEmpty() for type V4s
makeEmpty() for type V4i
makeEmpty() for type V4i64
makeEmpty() for type V4f
makeEmpty() for type V4d
makeInfinite() for type V2s
makeInfinite() for type V2i
makeInfinite() for type V2i64
makeInfinite() for type V2f
makeInfinite() for type V2d
makeInfinite() for type V3s
makeInfinite() for type V3i
makeInfinite() for type V3i64
makeInfinite() for type V3f
makeInfinite() for type V3d
makeInfinite() for type V4s
makeInfinite() for type V4i
makeInfinite() for type V4i64
makeInfinite() for type V4f
makeInfinite() for type V4d
extendBy() point for type V2s
extendBy() point for type V2i
extendBy() point for type V2i64
extendBy() point for type V2f
ImathTest: /builddir/Imath-3.1.5/src/ImathTest/testBox.cpp:257: void {anonymous}::testExtendByPoint(const char*) [with T = Imath_3_1::Vec2<float>]: Assertion `b.min == min && b.max == max' failed.
