Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Compilation errors with release candidate for BH package 1.75.0 #37

Open
eddelbuettel opened this issue Dec 13, 2020 · 4 comments
Open

Comments

@eddelbuettel
Copy link

As mentioned in issue 76 at the BH repo, its new version creates compilation issues with package TDA. I have not had time to really dive into this, but this may be caused by TDA also including some Boost headers to that we may get a conflict between different versions. The container library is included in BH so you could skip that; dealing with serialization may be trickier but an update to the current version may help.

Please reach out if you have any questions, and a big Thank You! for maintaining TDA on CRAN.

@jkim82133
Copy link
Collaborator

Hello, I am really sorry and realized few days ago that this email / issue thread was referencing TDA.

As you suggested, I removed duplicate container/detail/copy_move_algo.hpp from our own copy in TDA. Then TDA compiled fine, but after submitting it to CRAN, I got warning messages:

https://win-builder.r-project.org/incoming_pretest/TDA_1.7_20210126_170540/Windows/00check.log
https://win-builder.r-project.org/incoming_pretest/TDA_1.7_20210126_170540/Debian/00check.log

The warning messages are basically complaining about memmove function [-Wclass-memaccess]. I had the same problem before. Back then, I fixed that warning by modifying the first argument of memmove in copy_move_algo.hpp with applying static_cast<void *>:
std::memmove(dest_raw, beg_raw, sizeof(value_type)*n);
-> std::memmove(static_cast<void *>(dest_raw), beg_raw, sizeof(value_type)*n);
std::memmove(boost::movelib::iterator_to_raw_pointer(r), boost::movelib::iterator_to_raw_pointer(f), sizeof(value_type)*n);
-> std::memmove(static_cast<void *>(boost::movelib::iterator_to_raw_pointer(r)), boost::movelib::iterator_to_raw_pointer(f), sizeof(value_type)*n);
std::memmove(boost::movelib::iterator_to_raw_pointer(r), boost::movelib::iterator_to_raw_pointer(f), sizeof(value_type)*n);
-> std::memmove(static_cast<void *>(boost::movelib::iterator_to_raw_pointer(r)), boost::movelib::iterator_to_raw_pointer(f), sizeof(value_type)*n);
std::memmove((boost::movelib::iterator_to_raw_pointer)(r), (boost::movelib::iterator_to_raw_pointer)(f), sizeof(value_type)*n);
-> std::memmove(static_cast<void *>((boost::movelib::iterator_to_raw_pointer)(r)), (boost::movelib::iterator_to_raw_pointer)(f), sizeof(value_type)*n);

At that time, I modified boost library by myself since as I understand, the warning message is coming from the boost library itself. Is my understanding correct? Or is it something to be fixed on TDA side?

Sorry again and thank you!

@eddelbuettel
Copy link
Author

Hello, I am really sorry and realized few days ago that this email / issue thread was referencing TDA.

Well I opened it at your repo which commonly signals it concerns your code. Sorry if that wasn't clear. I believe I also emailed you in December just like the maintainers of the other packages.

Re the warnings: not sure. It may suggest a different initalization: _... with no trivial copy-assignment; use copy-assignment or copy-initialization instead _ What are the lines in your code that trigger this?

@eddelbuettel
Copy link
Author

eddelbuettel commented Jan 27, 2021

I think I just got one quieted down. In essence instead of

 MyType  x = MyType(a, b, c);

do

     MyType  x(a, b, c);

My editor reindent so apologies for the but basically this:

diff --git a/src/diag.cpp b/src/diag.cpp
index ed3f894..e5ec98c 100644
--- a/src/diag.cpp
+++ b/src/diag.cpp
@@ -402,9 +402,7 @@ Rcpp::List AlphaShapeFiltration(
 
   Rcpp::NumericMatrix coordinates;
 
-  Gudhi::Simplex_tree<> smplxTree =
-      AlphaShapeFiltrationGudhi< Gudhi::Simplex_tree<> >(
-          X, printProgress, Rprintf, coordinates);
+  Gudhi::Simplex_tree<> smplxTree(AlphaShapeFiltrationGudhi< Gudhi::Simplex_tree<> >(X, printProgress, Rprintf, coordinates));
   Rcpp::List filtration =
       filtrationGudhiToRcpp< Rcpp::List, Rcpp::NumericVector >(smplxTree);
   filtration.push_back(coordinates);
@@ -507,4 +505,4 @@ Rcpp::List AlphaComplexDiag(
       concatStlToRcpp< Rcpp::NumericMatrix >(persDgm, true, 3),
       concatStlToRcpp< Rcpp::NumericMatrix >(persLoc, false, 2),
       StlToRcppMatrixList< Rcpp::List, Rcpp::NumericMatrix >(persCycle));
-}
\ No newline at end of file
+}

Edit: Or maybe not. It's a bit hard to tell as there is as usual a fair amount of noise Boost and all that.

@jkim82133
Copy link
Collaborator

Sorry, I needed a quite amount of time to delve into this thing.
As I understand, the [-Wclass-memaccess] warning message is to warn when the raw memory function call might bypass the class's constructor or copy assignment or etc. ( https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html ).

And for the case of R package TDA, this warning message comes when the container boost::flat_map is used with a class with nontrivial constructor or nontrivial copy assignment. And this warning message comes when 1) constructor / copy assignment is called, or 2) memory access function from the container boost::flat_map is called such as flat_map::emplace_hint() or flat_map::erase().

  1. can be avoided when copy assignment is avoided such as your suggestion, so I tried this (which needed to restructure my code) and silenced many warning messages. But still, at some part of the algorithm calling emplace_hint() or erase() is inevitable. In boost library, these kind of memory functions are implemented with raw memory function memmove() (which is located in boost/container/detail/copy_move_algo.hpp ). For example, in the corresponding install.out ( https://win-builder.r-project.org/incoming_pretest/TDA_1.7.1_20210208_173610/Debian/00install.out , https://win-builder.r-project.org/incoming_pretest/TDA_1.7.1_20210208_173610/Windows/00install.out ), the [-Wclass-memaccess] message is coming from calling flat_map::emplace_hint() from boost library with the Gudhi class Gudhi::Simplex_tree_node_explicit_storage<Gudhi::Simplex_tree<> >, which has nontrivial copy constructor.

So I think it is neither the TDA package nor BH package to be fixed since the boost library is implemented in accessing raw memory functions. Rather, the [-Wclass-memaccess] warning message should be considered a false positive. So I am trying to convince CRAN package maintainers.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants