06:14 < sakshamb189[m]> Hey himanshu_pathak I think it looks mostly good.

06:14 < sakshamb189[m]> I see a few issues where you have capitalised words that should not be capitalised. This Summer -> This summer

06:14 < sakshamb189[m]> Artificial Neural network -> artificial neural network

11:03 < shrit[m]> OK I am here rcurtin

12:32 < rcurtin> shrit[m]: ok, maybe we can hop on zoom at 1300 UTC (half an hour from now)?

12:35 < shrit[m]> Perfect

14:23 < shrit[m]> rcurtin: I think we are getting the same error right? https://dev.azure.com/mlpack/mlpack/_build/results?buildId=4182&view=logs&j=24d3abe3-ef0b-5deb-3aab-64d839de2c3c&t=8be94158-0791-5881-12e7-451b62b18296&l=823

14:47 < rcurtin> shrit[m]: looks like this one is the same type of error but in a different place

14:48 < rcurtin> I followed the backtrace to cover_tree_impl.hpp

14:50 < rcurtin> looks like CoverTree also holds a const MatType*; I guess the same workaround will be needed

14:50 < shrit[m]> ok, I wil apply the same modification and see what happen

15:13 < rcurtin> you may want to do it locally in a docker container so that you don't have to wait for azure

15:14 < shrit[m]> I am starting to do this extact right now as you are saying this

15:14 < rcurtin> grepping for 'const MatType*' indicates a handful of classes that will probably need the same change:

15:14 < rcurtin> ./mlpack/core/tree/rectangle_tree/rectangle_tree.hpp: const MatType* dataset;

15:14 < rcurtin> ./mlpack/core/tree/cover_tree/cover_tree.hpp: const MatType* dataset;

15:14 < rcurtin> ./mlpack/core/tree/spill_tree/spill_tree.hpp: const MatType* dataset;

15:14 < rcurtin> ./mlpack/core/tree/binary_space_tree/vantage_point_split.hpp: const MetricType* metric;

15:15 < rcurtin> (I don't know if vantage_point_split has any serialize() function though, if not, that one can be ignored

15:15 < rcurtin> ./mlpack/methods/fastmks/fastmks.hpp: const MatType* referenceSet;

15:15 < rcurtin> (that one is already handled)

15:15 < rcurtin> ./mlpack/methods/rann/ra_search.hpp: const MatType* referenceSet;

15:15 < rcurtin> ./mlpack/methods/amf/termination_policies/simple_tolerance_termination.hpp: const MatType* V;

15:15 < rcurtin> (not sure if that class is serialized)

15:15 < rcurtin> ./mlpack/methods/range_search/range_search.hpp: const MatType* referenceSet;

15:15 < rcurtin> ./mlpack/methods/neighbor_search/neighbor_search.hpp: const MatType* referenceSet;

15:15 < rcurtin> you'll have to look through the serialize() implementations for each of those classes to see if the const pointer is being serialized, and the workaround might look a little different in each case

15:22 < shrit[m]> rcurtin: just a newbie docker question, how to get to install ubuntu 16. I have only ubunut 20 available in the official repo

15:25 < zoq> shrit[m]: docker run -it ubuntu:trusty /bin/bash

15:25 < zoq> shrit[m]: Should download the image.

15:25 < zoq> shrit[m]: Ahh, you said, 16.04

15:25 < zoq> shrit[m]: So ubuntu:xenial

15:26 < shrit[m]> OK, just have to specify the code name with

15:26 < shrit[m]> 👍️

15:26 < zoq> yes, here are the tags: https://hub.docker.com/_/ubuntu

15:26 < zoq> https://hub.docker.com/_/ubuntu?tab=tags

15:27 < rcurtin> yeah, exactly; 16.04 has cereal 1.1, and 18.04 has cereal 1.2

15:27 < rcurtin> in the best case, the workarounds we implement will work for 1.1, but if they don't, we can try 1.2 as a minimum version and see if we can make that work

15:27 < rcurtin> if not, then we'll have to do CMake + an auto-downloader for cereal 1.3

15:27 < rcurtin> luckily, cereal is header-only like ensmallen, so auto-downloaders aren't *too* scary

15:28 < zoq> rcurtin: Or just include the header?

15:29 < zoq> I guess there is no real benefit from downloading, if we download a specific version?

15:31 < rcurtin> zoq: cereal is kind of big, I'd prefer not to keep it in our repository (catch, for instance, is much smaller)

15:31 < rcurtin> the only benefit I think would be keeping the repository a bit smaller

15:31 < zoq> I see, yeah haven't thought about size.

15:31 < rcurtin> like the ensmallen auto-downloader, the intention would be to remove the autodownloader when a new enough version is available in major distros to install from the package manager

15:32 < zoq> right, makes sense

15:50 < jenkins-mlpack2> Yippee, build fixed!

15:50 < jenkins-mlpack2> Project mlpack - git commit test build #520: FIXED in 57 min: http://ci.mlpack.org/job/mlpack%20-%20git%20commit%20test/520/

18:39 < rcurtin> the R bindings are basically ready to merge: https://github.com/mlpack/mlpack/pull/2556

18:39 < rcurtin> it's a big PR, so I wanted to ask if anyone wants me to wait so they can review it before merging

19:17 < zoq> rcurtin: If, one more day doesn't matter, I'll take general look over the PR today/tomorrow.

19:18 < rcurtin> sure, I don't think there is a giant hurry :)

19:18 < rcurtin> I was hoping to release 3.4.0 once that is merged, and also once #2566 is merged (by the way I did fix the issue you pointed out there, if you have time to take another look :))

19:20 < zoq> rcurtin: Ohh, nice, yeah, will take a look at #2566 first, since it's much less code :)

19:20 < rcurtin> hehe :)