Closed yixuan closed 6 years ago
Awesome. I'll start a rev.dep run in a moment.
Looks good. I ran one, and saw no compilation issue. There were a few tests failing where packages had Suggests: yet required the package present---I will skip those packages in a second run.
One test failure (Cyclops) which gets the same failure under RcppEigen 0.3.3.3.1, so no regression.
Hu @yixuan should we care about this? Logs from r-devel on 0.3.3.4.0: https://win-builder.r-project.org/bWF1JR2EGjK6/00check.log
I guess we may be able to argue the long long
away ("them, not us"). Should we deal with the pragma issue?
My bad. I accidentally overwrote this commit: https://github.com/RcppCore/RcppEigen/commit/a56f72dc1cf42aaf8c9f4d5beb466ab46fa89f1e. Could you make a quick fix? I think I should report it to Eigen so that next time we do not need to make a manual patch.
No worries. You can either commit to the same PR (I think) or just open a new one (maybe simpler).
I just reapplied https://github.com/RcppCore/RcppEigen/commit/a56f72dc1cf42aaf8c9f4d5beb466ab46fa89f1e as d934b12c12886fece65b7bc57030e61c735566a3 but am a little worried about the remainder of the build log:
* checking pragmas in C/C++ headers and code ... WARNING
File which contains pragma(s) suppressing important diagnostics:
'inst/include/Eigen/src/Core/arch/SSE/Complex.h'
File which contains pragma(s) suppressing diagnostics:
'inst/include/Eigen/src/Core/util/DisableStupidWarnings.h'
* checking compiled code ... OK
We may have to comment the pragmas out. :-/
Done in 6247fd632ad6ee78daffa0acff9ff1af8d7aaaa5 and I will run another set of reverse depends tests later; easy enough now thanks to prrd.
Thanks for the help! I'll document them in my notes and remember to apply these manual patches next time. prrd
looks great!
Shall we, just for our reference, create a recursive diff of "our" Eigen 3.3.4 versus upstream? We could stick it into a top-level directory excluded via .Rbuildignore
.
Great. Will do that later today.
BTW suppressing the #pragma
makes even R CMD INSTALL RcppEigen
extremely "noisy" :-/
Yeah, I see that... Even the 0.3.3.3.1 package now has warnings under R-devel: https://cran.r-project.org/web/checks/check_results_RcppEigen.html.
Yes, those checks are new-ish. (I run one robot diff'ing CRAN Policy and tweeting / committing changes. I think this may have been added last November or December.)
I cherry-picked your commit with the diff to Eigen 3.3.4. Minor change to directory layout but nothing substantial. Thanks for setting that up.
Thank you Dirk!
@yixuan if I may ask how do you decide what to include or how you figure out what is in each version? I saw this previous pull request on the Eigen bitbucket here that contains additional changes (among a few additional small changes also in the github mirror) I need for some of my work. I have made the updates in my own fork of RcppEigen but I don't know how you want to manage pull requests.
@yixuan Do you know how else I can get in touch with the Eigen developers? I tried subscribing to the mailing list but I have yet to hear anything back over the last two weeks. Wasn't sure if there was another way to reach out.
@cdeterman Gael has a public email gael.guennebaud@gmail.com, but from the commit log of Eigen I guess he was tied up by something else and didn't get the time for Eigen development at this moment.
This PR updates the included Eigen to 3.3.4, and also applies some upstream patches used to fix #48 .
@eddelbuettel I haven't got the time to run a reverse-dependence check for this PR. Could you run it on the automated server and show me the report? I'd like to look into the cause of potential failures as usual. Thanks!