Update multiply

[bjorgve]

Making vampyr work with current master of mrcpp

[bjorgve]

I'm not sure why the CIs are failing. Everything works smoothly on Woolf. Do you know whats wrong @robertodr?

[stigrj]

We recently had to update the .travis.yml config in mrchem, looks like the same issue. Circle is complaining about a public key in the clone, don't know what that is about.

[robertodr]

You might also want to update the version of MRCPP that is fetched (look inexternal/CMakeLists.txt)

[bjorgve]

If I'm not wrong it looks like it fetches mrcpp/master. But we should probably fetch a release by default

[robertodr]

Indeed!

[bjorgve]

@stigrj I don't see anything different between mrchem/travis.yml and vampyr/travis.yml (apart from sudo: false). Which PR are you thinking about?

[robertodr]

It's on MRChem's master branch: https://github.com/MRChemSoft/mrchem/blob/master/.travis.yml

[robertodr]

Not sure what's wrong with CircleCI here... Something about access to the repo, even thought it downloads anonymously...

[robertodr]

I've updated the Docker image, so you should now use: - image: quay.io/metamr/circleci_ubuntu-18.04-conda:41efbc8 in config.yml However, this doesn't fix the problem either, maybe @stigrj needs to reauthorize CircleCI or something...
`

[stigrj]

I can reproduce the aborted child error from Travis in debug

[stigrj]

I get the same abortion with master using mrcpp-1.2.0 in debug, so it's probably not related to this PR

[stigrj]

Should vampyr always point to master, or should we pin it to versions?

[stigrj]

I get double free or corruption errors for the Eigen::MatrixXds in ScalingBasis when compiled in debug through vampyr. If I make them pointers and let them leak, then the tests pass.

[robertodr]

I'd say we should pin the dependency. I'll try running through the debugger later to see what's going on with the double free skullduggery...

[robertodr]

The debugger did not provide illumination. I've compiled on sandel and the double free occurs when doing phi_tree = vp.vampyr3d.FunctionTree(MRA) It could be the binding is not 100% correct when it comes to inheritance... I'm looking into it, but I'm not sure I can fix it quickly.

[stigrj]

I narrowed it down to a single line in the simplest case

basis = vp.InterpolatingBasis(5)

So just initialization and destruction of a ScalingBasis, in particular one of its three Eigen::MatrixXd members.

[robertodr]

Hmmm I removed the destructor in the base class in MRCPP and that python line doesn't fail for me.

…

On Mon, Sep 21, 2020, 09:25 Stig Rune Jensen ***@***.***> wrote: I narrowed it down to a single line in the simplest case basis = vp.InterpolatingBasis(5) So just initialization and destruction of a ScalingBasis, in particular one of its three Eigen::MatrixXd members. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#34 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA4JOENFYFNNGAFJN7COTOLSG35VRANCNFSM4QUYXR4A> .

[codecov]

Codecov Report

Merging #34 into master will decrease coverage by 0.07%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #34      +/-   ##
==========================================
- Coverage   85.81%   85.73%   -0.08%     
==========================================
  Files          30       30              
  Lines         578      575       -3     
==========================================
- Hits          496      493       -3     
  Misses         82       82              

Impacted Files	Coverage Δ
vampyr/math/math.h	100.00% <100.00%> (ø)
vampyr/functions/GaussExp.h	82.35% <0.00%> (-2.65%)	⬇️
vampyr/functions/GaussFunc.h	72.72% <0.00%> (-2.28%)	⬇️
vampyr/operators/BSOperator.h	75.00% <0.00%> (ø)
vampyr/operators/PHOperator.h	75.00% <0.00%> (ø)
vampyr/operators/ABGVOperator.h	100.00% <0.00%> (ø)
vampyr/operators/PoissonOperator.h	100.00% <0.00%> (ø)
vampyr/operators/HelmholtzOperator.h	100.00% <0.00%> (ø)
vampyr/operators/IdentityConvolution.h	33.33% <0.00%> (ø)

---

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 89ac77a...ae5fd31. Read the comment docs.

[stigrj]

So there was a missing -march=native flag. I'm deja vu-ing

[robertodr]

The agony! That flag should be exported as part of the MRCPP target then. Or removed altogether?

[robertodr]

Can you make codecov happy?

[robertodr]

FYI, MRChemSoft/mrcpp#147 does the export of the flag with the target.

[robertodr]

What a trip this "simple" change was 😸 I'll push some other changes to the bindings as follow-up.

[robertodr]

Now that all bots are happy, we can happily merge. Thanks @bjorgve!

[bjorgve]

Wow you guys did allot of work here! Thanks!