Boost logo

Boost :

Subject: Re: [boost] [1.41.0] [phoenix] [accumulators] [maths] [ublas] Beta 1 release candidate
From: Phil Richards (news_at_[hidden])
Date: 2009-11-04 15:41:07

On Tue, 2009-11-03 at 20:10 +0000, Phil Richards wrote:
> The tests are now running - I'll check over the results and post if
> there are any unexpected failures, but that isn't going to be anytime
> soon :-)

My test run has completed (gcc 4.4.1, Ubuntu 9.10, x86, release - no MPI
- I've not managed to get bjam to recognise that MPI is available yet -
more configuration problems on my part, I suspect).

Every test in spirit/phoenix was marked as a failure: looking at the
failure report, none of them actually failed. (All had a success report
for the test.)

There are a large number of warnings (from the compiler warnings already
posted by myself and John Maddock), but there also seem to be quite a
lot of errors in maths related tests.

I suspect that most of these problems are purely related to the fact
that this is a release build with optimization turned on.

I wouldn't have thought that any of these are ones that would affect the

For example, in accumulators:

Run output:
Running 1 test case...
pot_quantile.cpp(82): error in "test_stat": difference{1.0227%} between quantile(acc5, quantile_probability = 0.999){6.9786481796869042} and 6.908{6.9080000000000004} exceeds 1%

*** 1 failure detected in test suite "Master Test Suite"

Run output:
Running 1 test case...
weighted_tail_mean.cpp(61): error in "test_stat": difference{1.11208%} between non_coherent_weighted_tail_mean(acc1, quantile_probability = 0.025){0.0123609897219755} and 0.0125{0.012500000000000001} exceeds 1%
weighted_tail_mean.cpp(62): error in "test_stat": difference{2.44744%} between non_coherent_weighted_tail_mean(acc1, quantile_probability = 0.01){0.0048776278773278257} and 0.005{0.0050000000000000001} exceeds 1%

*** 2 failures detected in test suite "Master Test Suite"


(Plus more)

And in maths:

Run output:
Running 1 test case...
Tolerance for type f is 0.01 %
test_normal.cpp(233): error in "test_main_caller( argc, argv )": difference{1.09314e-05%} between hazard(dist, x){0.00425984897} and pdf(dist, x) / cdf(complement(dist, x)){0.00425984943} exceeds 5.96046448e-07%
Tolerance for type d is 0.01 %
Tolerance for type e is 0.01 %
Tolerance for type N5boost4math8concepts12real_conceptE is 0.01 %

*** 1 failure detected in test suite "Test Program"


(Plus more)

And in numeric/ublas:

Run output:
factorization completed: 1
LU factors are correct: 0
permutation is correct: 1
inverse is correct: 0


Hope this is of some use,

Phil Richards, <news_at_[hidden]>

Boost list run by bdawes at, gregod at, cpdaniel at, john at