Problem after changing from hermes 2.0 to hermes 3.0

[Jens Niegemann]

Hi,
I recently decided to check out the new version 3.0 of hermes and tried to adapt my old mode solver to the new version. This seemed fairly straight-forward and I appreciate the use of smart pointer all over the place. So, basically all my changes were to replace pointers by the corresponding smart pointers. No other changes were made.

Unfortunately, the version ported to hermes 3.0 results in wrong matrices. Now, I am not sure if this is my fault because something changed (or was tolerated by the previous version) but the problem seems to  be concentrated on systems of equations. If I dump the matrices produced by 2.0 and 3.0, it seems that 3.0 does not properly offset the matrix entries for the individual components of the system of equations. I attached two spy-plots, one from the old version and one from the new version. It clearly shows that in 3.0 the matrix blocks all seem to start at (1,1).

Now, my question is whether this could be a bug in hermes or if the usage for systems of equations somehow changed (I tried to figure this out but could not find anything) and I need to fix my code?

Best regards,
Jens

[Lukas Korous]

Hi Jens,

I am sorry for the late reply (vacation) - sorry to hear that you are
having problems - could you send me the code of yours? If you had it
on github that would be great (if I could clone -> fix (either hermes
or your code) -> push).

Best,
Lukas

> --
> You received this message because you are subscribed to the Google Groups
> "hermes2d" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to hermes2d+u...@googlegroups.com.
> To post to this group, send email to herm...@googlegroups.com.
> Visit this group at http://groups.google.com/group/hermes2d.
> For more options, visit https://groups.google.com/d/optout.



--
Lukáš Korous