search

boutproject / BOUT-dev

BOUT++: Plasma fluid finite-difference simulation code in curvilinear coordinate systems
http://boutproject.github.io/
GNU Lesser General Public License v3.0
179 stars 93 forks source link

Fix metrics in Laplace inversion #2914

Closed bendudson closed 1 month ago

bendudson commented 1 month ago

Was using metric tensors at wrong cell index when NXPE > 1, shifted by 2 cells in X.

Only matters when setting boundaries to specific gradients.

ZedThree commented 1 month ago

@dschwoerer Looks like there's an issue with the bout++ COPR:


+ dnf5 -y builddep bout++
Updating and loading repositories:
 Fedora rawhide openh264 (From Cisco) - 100% | 637.3 KiB/s |   1.5 MiB |  00m02s

...

Problem 1: package sundials-openmpi-devel-6.7.0-1.fc41.x86_64 requires sundials-openmpi(x86-64) = 6.7.0-1.fc41, but none of the providers can be installed
  - conflicting requests
  - nothing provides libsuperlumt64_d.so.3.1()(64bit) needed by sundials-openmpi-6.7.0-1.fc41.x86_64
 Problem 2: conflicting requests
  - package sundials-mpich-devel-6.7.0-1.fc41.i686 requires sundials-mpich(x86-32) = 6.7.0-1.fc41, but none of the providers can be installed
  - package sundials-mpich-devel-6.7.0-1.fc41.x86_64 requires sundials-mpich(x86-64) = 6.7.0-1.fc41, but none of the providers can be installed
  - nothing provides libsuperlumt_d.so.3.1 needed by sundials-mpich-6.7.0-1.fc41.i686
  - nothing provides libsuperlumt64_d.so.3.1()(64bit) needed by sundials-mpich-6.7.0-1.fc41.x86_64
 Problem 3: conflicting requests
  - package sundials-devel-6.7.0-1.fc41.i686 requires sundials(x86-32) = 6.7.0-1.fc41, but none of the providers can be installed
  - package sundials-devel-6.7.0-1.fc41.x86_64 requires sundials(x86-64) = 6.7.0-1.fc41, but none of the providers can be installed
  - nothing provides libsuperlumt_d.so.3.1 needed by sundials-6.7.0-1.fc41.i686
  - nothing provides libsuperlumt64_d.so.3.1()(64bit) needed by sundials-6.7.0-1.fc41.x86_64 ```
ZedThree commented 1 month ago

Test failures are unrelated. This bug is perhaps more evidence we could really do with a much more comprehensive laplacian test suite -- easier said than done perhaps!