Tests failures for summation algorithms. math-functions-0.1.5.2

[peti]

Citing from http://hydra.nixos.org/build/25138672/nixlog/1/raw:

Running 1 test suites...
Test suite tests: RUNNING...
Special functions:
  Gamma(x+1) = x*Gamma(x) [logGamma]: [OK, passed 100 tests]
  Gamma(x+1) = x*Gamma(x) [logGammaL]: [OK, passed 100 tests]
  gamma(1,x) = 1 - exp(-x): [OK, passed 100 tests]
  0 <= gamma <= 1: [OK, passed 100 tests]
  0 <= I[B] <= 1: [OK, passed 100 tests]
  invErfc = erfc^-1: [OK, passed 100 tests]
  invErf  = erf^-1: [OK, passed 100 tests]
  Factorial is expected to be precise at 1e-15 level: [OK]
  Log factorial is expected to be precise at 1e-15 level: [OK]
  logGamma is expected to be precise at 1e-9 level [integer points]: [OK]
  logGamma is expected to be precise at 1e-9 level [fractional points]: [OK]
  logGammaL is expected to be precise at 1e-15 level: [OK]
  logGammaL is expected to be precise at 1e-10 level [fractional points]: [OK]
  logBeta is expected to be precise at 1e-6 level: [OK]
  digamma is expected to be precise at 1e-14 [integers]: [OK]
  digamma is expected to be precise at 1e-12: [OK]
  incompleteBeta is expected to be precise at 1e-8 level: [OK]
  incompleteBeta with p > 3000 and q > 3000: [OK]
  choose is expected to precise at 1e-12 level: [OK]
  Self-test: 0 <= range01 <= 1: [OK, passed 100 tests]
Chebyshev polynomials:
  Chebyshev 0: [OK, passed 100 tests]
Summation:
  Kahan:
    t_sum_shifted: [OK, passed 100 tests]
  KBN:
    t_sum: [Failed]
*** Failed! Falsifiable (after 10 tests and 3292 shrinks): 
[1.12e-321,4.450748902134673e-308,1.3349861871991834e-307]
(used seed 505504302)
    t_sum_error: [OK, passed 100 tests]
    t_sum_shifted: [Failed]
*** Failed! Falsifiable (after 4 tests and 34 shrinks): 
[9.765630238698301e-4,0.0,0.0]
(used seed -111987417)
  KB2:
    t_sum: [Failed]
*** Failed! Falsifiable (after 11 tests and 3202 shrinks): 
[5.0e-324,4.450329066239099e-308,1.3350264904076728e-307]
(used seed -1471674427)
    t_sum_error: [OK, passed 100 tests]
    t_sum_shifted: [Failed]
*** Failed! Falsifiable (after 7 tests and 28 shrinks): 
[9.765628630434347e-4,0.0,0.0]
(used seed -815096692)

         Properties   Test Cases   Total       
 Passed  12           12           24          
 Failed  4            0            4           
 Total   16           12           28          
Test suite tests: FAIL
Test suite logged to: dist/test/math-functions-0.1.5.2-tests.log
0 of 1 test suites (0 of 1 test cases) passed.

I believe that the i686 architecture may play a part in this issue; we don't see those failures on 64 bit platforms.

[Shimuuar]

I think it was fixed in ff01fcc and uploaded to hackage as 0.1.5.3