|
|
Boost Testing : |
From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2005-12-16 03:03:50
Martin Wille writes:
> Aleksey Gurtovoy wrote:
>
>> Somebody else's results prevented generation of the Boost-wide
>> XML. We'll investigate it further ASAP, meanwhile I've deleted the
>> offending archive, so unless there are more showstoppers lurking
>> ahead, the new reports should be on the web within a couple of hours.
>
> It didn't work, apparently.
Please check again -- we've just fixed it.
> Is there anything wrong with my archive?
There was; or, to be precise, not with your archive per se, but one of
the tests, "libs/math/test/complex_test.cpp". When compiled with
'gcc-3.2.3-linux', the executable's output looks like this:
<run result="fail" timestamp="2005-12-12 10:43:49 UTC">
Running 1 test case...
Running complex trig sanity checks for type float.
../libs/math/test/complex_test.cpp(85): error in "test_main_caller( argc, argv )": Error in result exceeded permitted limit of 50 (actual relative error was 1.18927e+07e). Found (0.83373,-0.988898) expected (-1,-1)
../libs/math/test/complex_test.cpp(85): error in "test_main_caller( argc, argv )": Error in result exceeded permitted limit of 5 (actual relative error was 1.63701e+07e). Found (1.29846,0.634964) expected (-1,-1)
../libs/math/test/complex_test.cpp(85): error in "test_main_caller( argc, argv )": Error in result exceeded permitted limit of 50 (actual relative error was 1.18926e+07e). Found (0.83373,-0.988898) expected (-1,-0.984375)
../libs/math/test/complex_test.cpp(85): error in "test_main_caller( argc, argv )": Error in result exceeded permitted limit of 5 (actual relative error was 1.63177e+07e). Found (1.29846,0.634964) expected (-1,-0.984375)
../libs/math/test/complex_test.cpp(85): error in "test_main_caller( argc, argv )": Error in result exceeded permitted limit of 50 (actual relative error was 1.18932e+07e). Found (0.83373,-0.988898) expected (-1,-0.96875)
...
[ Ad infinitum, 2,171,417 lines, 468 Mb of text! ]
Apparently, 'xsltproc' isn't well-prepared to deal with this much data
within a single element. The other runner's results that I've
mentioned earlier ("FFT-altix") had the same issue.
We are truncating everything past 64 Kb now, see
http://tinyurl.com/9kbx5
(http://engineering.meta-comm.com/boost-regression/CVS-HEAD/developer/output/Martin%20Wille-bin-boost-libs-math-test-complex_test-test-gcc-3-2-3-linux-debug.html).
The test still needs to be looked at, though.
-- Aleksey Gurtovoy MetaCommunications Engineering