On Sat, Jun 7, 2008 at 9:15 AM, Thomas Klimpel <Thomas.Klimpel@synopsys.com> wrote:

Hello Jeremy,

> I hate to keep filling your inboxes, but does anyone have any ideas?

Don't worry about filling our inboxes, we subscribed voluntarily to boost-ublas.
Your mistake is not in hseqr.hpp, but in test_uBlas.cpp. You use a real matrix Z in the complex case (line 117).

I was a bit surprised to see that test_uBlas.cpp compiled, linked and executed "successful" after fixing this. I thought that line 183 and line 220 in hseqr.hpp are wrong, because you pass a pointer to work_temp instead of a pointer to work.

Thanks! that did the trick. I'm glad you also caught the problems on lines 183 and 220, even though it compiled successfully, it was wrong.

So I have gone about as far as I can. I'm going to run a few tests to make sure it gives what I expect. I know there isn't a lot of error checking; what there is, is copied from some of the other libraries.

Thanks everyone for the help,

Jeremy