trixi-framework / HOHQMesh

High Order Hex-Quad Mesh (HOHQMesh) package to automatically generate all-quadrilateral meshes with high order boundary information.

https://trixi-framework.github.io/HOHQMesh

Other

52 stars 8 forks source link

Fix Windows CI errors #39

Closed sloede closed 2 years ago

sloede commented 2 years ago

Possible fixes for the build error on Windows for main:

~~User previous GH environment Windows 2019~~ Does not work at all (msys2 not working, see 336df420df41070b756dbf0c70f349de76bf476a)
~~Downgrade to gfortran v11.3~~ Not possible using msys2, see, e.g., here and 95b4cdcc2e3a4ffffd86f274d2cd2589178c56d7
~~Reproduce error with GCC 12.1 under Linux/macOS and fix it there~~ Error not reproducible on roci
Compile with debug symbols for more useful error message NOT TESTED
Run tests through gdb for more useful error message NOT TESTED
Use gcc-fortran package in msys2 ✅

codecov[bot] commented 2 years ago

Codecov Report

Merging #39 (aeda203) into main (8a539e9) will not change coverage. The diff coverage is n/a.

@@           Coverage Diff           @@
##             main      #39   +/-   ##
=======================================
  Coverage   67.50%   67.50%           
=======================================
  Files          68       68           
  Lines       10200    10200           
=======================================
  Hits         6885     6885           
  Misses       3315     3315

Flag	Coverage Δ
unittests	`67.50% <ø> (ø)`

Flags with carried forward coverage won't be shown. Click here to find out more.

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 8a539e9...aeda203. Read the comment docs.

sloede commented 2 years ago

@DavidAKopriva @andrewwinters5000 is this good to merge?

andrewwinters5000 commented 2 years ago

@sloede Maybe, I have no way to verify / reproduce these issues without a Windows machine. From some of the detective work from @fluidnumerics-joe I know that @DavidAKopriva has fixed some of the uninitialized variables in #42

DavidAKopriva commented 2 years ago

Yes, I have another PR with a bunch of uninitialized variables fixes. Waiting for the test, but if it is OK, we will not need this branch.

sloede commented 2 years ago

My suggestion is to merge this branch immediately, such that CI testing works again for main and in other PRs.

DavidAKopriva commented 2 years ago

Doesn't this version use an older gfortran? Windows now works in the other branch with the newer fortran. Sooner or later we will want to use the latest.

sloede commented 2 years ago

Doesn't this version use an older gfortran? Windows now works in the other branch with the newer fortran. Sooner or later we will want to use the latest.

@DavidAKopriva Sorry, I posted my comment before I read that you had answered already. Of course, if you have a proper fix in another PR that makes Windows work again, that should be the one to be merged. Feel free to request a review on #42 if I can help. And yes, once #42 is merged, we should close this one.

DavidAKopriva commented 2 years ago

Oh, good. I was just a little confused.

sloede commented 2 years ago

Superseded by #42.