00:31 < zoq> marcosirc: hm, it worked for me, can you point me to the commit?

00:33 < marcosirc> zoq: Hi! Sorry, I am not sure I understand.

00:33 < marcosirc> The last commit of the PR is: https://github.com/mlpack/mlpack/pull/747/commits/69804c65f972dfa71be1933ef5d49723baccfb00

00:33 < marcosirc> does it works for you?

00:34 < marcosirc> I removed the commit where I was including a lot of extra code to make many default template parameters explicit, because it didn't make a difference for appVeyor.

00:38 < zoq> I think the only change I made was: https://gist.github.com/zoq/7307fa7c72f9e8016dec3e696ae7e2d7 based on https://github.com/MarcosPividori/mlpack/commit/69804c65f972dfa71be1933ef5d49723baccfb00

00:39 < zoq> with that change I was able to build it on windows

00:40 < zoq> at least that was my impression, I'll go and check it again in a couple of hours

00:44 < marcosirc> zoq: Ok, thanks for your comments. I will change that a make a temporary commit to see what happens with appVeyor!

00:44 < marcosirc> *and

02:41 < marcosirc> zoq: that change made appVeyor succeed!

02:43 < marcosirc> Well, in fact, not succeed, now it prints a different error.

02:43 < marcosirc> "fatal error C1128: number of sections exceeded object file format limit: compile with /bigobj"

02:45 < marcosirc> it looks like the solution is to activate the compiler flag: /bigobj

08:42 < zoq> ah, okay, in fact I've always had to build with /bigobj.

16:40 < rcurtin> marcosirc: the spill tree PR finally built with appveyor! :)

16:54 < marcosirc> rcurtin: yeah!! :)

18:10 < marcosirc> rcurtin: I have made a PR including the implementation of a general greedy single tree traverser.

18:11 < marcosirc> I think it is ready to be merged. Both Travis and AppVeyor have succeeded.

18:16 < rcurtin> ok, I have a few PRs to look at first but I will look when I have a chance

18:17 < marcosirc> Ok, sure! I have a lot to do on SpillTree's PR.

18:39 < rcurtin> yeah, I guess I left a lot of comments, hopefully it is not too much work :)

20:14 < marcosirc> 'arma::linspace<vec>(n, m, 0)' should return an empty vector, shouldn't it?

20:14 < marcosirc> it looks like it is returning a vector with one element 'm'.

20:24 < rcurtin> interesting, it sounds like maybe that is an Armadillo bug

20:26 < rcurtin> I looked at the code for Armadillo 6.700.6, it is definitely coded that even if N = 0, it will return a vector of length 1

20:26 < rcurtin> I would suggest sending Conrad Sanderson an email asking if that is intended behavior

20:34 < marcosirc> rcurtin: Ok. Thanks. I found it doing some "extra" testing.

20:34 < marcosirc> I will send him an email.

20:35 < marcosirc> I was wondering why he doesn't provide the source code on github.

20:37 < rcurtin> the question has been asked for many years...

20:37 < rcurtin> I believe that he wants to keep his maintenance load low, so he does not want to put it on Github where people will file PRs and submit issues

20:38 < rcurtin> I disagree with his reasoning but in the end it's his library, so I am not in a position to influence his decision :)

20:46 < marcosirc> I understand... I think most people prefer to create an issue on github rather than writing an email.