Thanks for letting us know. However, we would be greatful if you provide some more information about these two crashes. Are you able to build the debug library and run these two praticular tests using the debugger? We have never encountered any crashes in the bessely() function so far.
BR,
/ediap
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I think the problem with bessely() function under MSVC++ is caused by naming differences of functions in <cmath>. Microsoft uses _y0(), _y1() and _yn(), whereas POSIX names are y0(), y1() and yn(). I will try to fix this, but without checking if it works because I use Linux/Unix platforms only.
BR,
/ediap
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thanks for you help with tracing this MSVC++ related bugs. I will let you know as soon as I update the sources, so you could check them under MSVC++.
Hope to write to you soon!
BR,
/ediap
PS. It seems that the email you provided is incorrect:
----- The following addresses had permanent fatal errors -----
<zhlei_ustc@gmail.com>
(reason: 550 5.7.1 No such user d2si442226nfe)
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I would like you to download (from the same link as previously), compile and test the modified IT++ sources under MSVC++ .NET.
I think that I fixed the problem of lgamma and gamma functions under MSVC++ .NET. The only thing to verify is if "signgam" variable is declared in <cmath> under MSVC++. If not, its declaration has to be added to itpp/base/scalfunc.cpp near the lgamma() function definition.
Hi,
I have tried all the test code in the \tests\ folder, there are two cases failed and cause the crash.
one is the bessel test. the bessely() will cause crash. others are ok.
one is the filter_design_test. I suspect it has memory leakage.
I am using ITpp3.10 with MSVC.NET and MKL8.0.
Alex
Dear Alex,
Thanks for letting us know. However, we would be greatful if you provide some more information about these two crashes. Are you able to build the debug library and run these two praticular tests using the debugger? We have never encountered any crashes in the bessely() function so far.
BR,
/ediap
Alex,
Thanks for your help in bugfixing IT++ under MSVC++ .NET. The fixes are now in Subversion repository and will be included in next releases of IT++.
/ediap
I think the problem with bessely() function under MSVC++ is caused by naming differences of functions in <cmath>. Microsoft uses _y0(), _y1() and _yn(), whereas POSIX names are y0(), y1() and yn(). I will try to fix this, but without checking if it works because I use Linux/Unix platforms only.
BR,
/ediap
OK. This problem should be fixed in Subversion repositories (trunk and branches/itpp-3-10).
I would be glad if you could check the following SVN snapshot package and let me know the status of the tests:
http://hydrus.et.put.poznan.pl/~ediap/download/itpp-3.10.1.tar.bz2
Thanks in advance!
/ediap
Hi Adam,
Thanks for you quick response.
I have checked the new version, unfortunately it doesn't work.
I traced it and found it crashes at jv.cpp line 467. there is an it_error("lgamma only defined for positive values for Visual C++"), and then crashes.
I would like help you work it out with my MSVC environment. When you have new version, just send the modified file to my email: zhlei_ustc@gmail.com
Alex
Hi Alex,
Thanks for you help with tracing this MSVC++ related bugs. I will let you know as soon as I update the sources, so you could check them under MSVC++.
Hope to write to you soon!
BR,
/ediap
PS. It seems that the email you provided is incorrect:
----- The following addresses had permanent fatal errors -----
<zhlei_ustc@gmail.com>
(reason: 550 5.7.1 No such user d2si442226nfe)
Dear Alex,
I would like you to download (from the same link as previously), compile and test the modified IT++ sources under MSVC++ .NET.
I think that I fixed the problem of lgamma and gamma functions under MSVC++ .NET. The only thing to verify is if "signgam" variable is declared in <cmath> under MSVC++. If not, its declaration has to be added to itpp/base/scalfunc.cpp near the lgamma() function definition.
Please find also a test program at the following location:
http://hydrus.et.put.poznan.pl/~ediap/download/gamma_test.cc
and compare the results with:
http://hydrus.et.put.poznan.pl/~ediap/download/gamma_test.ref
Thanks for your help!
/ediap
Hi Adam,
As you thought, the signgam is not declared in MSVC.NET. I uncomment it at scalfunc.cpp just before lgamma().
Then as we want, it works. The bessel tests are passed. Thank you.
I would like to help to test your future versions with my MSVC.NET environment if you want. You can contact me with my email.
Alex