Windows10下Rcpp::sourceCpp生成Debug DLL及相关配置问题咨询

问题背景 #

适用于Windows 10系统，使用Rcpp::sourceCpp编译C++函数并指定目录跟踪代码修改，遇到以下问题：

• 设置windowsDebugDLL=TRUE后仍生成优化代码，Debug构建未生效
• 不清楚链接时-lR对应的库文件，搜索libR.lib/libR.a/R.lib/R.a均未找到
• 编译所用的R Makefile位置，是否为临时生成后即被删除
• Windows下Makevars/Makevars.win的正确位置，放在用户目录.R文件夹无效，确认能否生成Debug版本DLL

编译输出详情 #

Rcpp::sourceCpp("sfgkanal.cpp", verbose=TRUE, rebuild=FALSE, cacheDir = getwd(), windowsDebugDLL=TRUE)
    Generated extern "C" functions 
--------------------------------------------------------


#include <Rcpp.h>
#ifdef RCPP_USE_GLOBAL_ROSTREAM
Rcpp::Rostream<true>&  Rcpp::Rcout = Rcpp::Rcpp_cout_get();
Rcpp::Rostream<false>& Rcpp::Rcerr = Rcpp::Rcpp_cerr_get();
#endif

// analyseSFGK
Rcpp::List analyseSFGK(Rcpp::List& x);
RcppExport SEXP sourceCpp_1_analyseSFGK(SEXP xSEXP) {
BEGIN_RCPP
    Rcpp::RObject rcpp_result_gen;
    Rcpp::RNGScope rcpp_rngScope_gen;
    Rcpp::traits::input_parameter< Rcpp::List& >::type x(xSEXP);
    rcpp_result_gen = Rcpp::wrap(analyseSFGK(x));
    return rcpp_result_gen;
END_RCPP
}
// libraryReady
Rcpp::String libraryReady();
RcppExport SEXP sourceCpp_1_libraryReady() {
BEGIN_RCPP
    Rcpp::RObject rcpp_result_gen;
    Rcpp::RNGScope rcpp_rngScope_gen;
    rcpp_result_gen = Rcpp::wrap(libraryReady());
    return rcpp_result_gen;
END_RCPP
}

Generated R functions 
-------------------------------------------------------

`.sourceCpp_1_DLLInfo` <- dyn.load('H:/jupyter-mktmicro-research/R/sfgk/sourceCpp-x86_64-w64-mingw32-1.0.9/sourcecpp_69f04dc58d9/sourceCpp_19.dll')

analyseSFGK <- Rcpp:::sourceCppFunction(function(x) {}, FALSE, `.sourceCpp_1_DLLInfo`, 'sourceCpp_1_analyseSFGK')
libraryReady <- Rcpp:::sourceCppFunction(function() {}, FALSE, `.sourceCpp_1_DLLInfo`, 'sourceCpp_1_libraryReady')

rm(`.sourceCpp_1_DLLInfo`)

Building shared library
--------------------------------------------------------

DIR: H:/jupyter-mktmicro-research/R/sfgk/sourceCpp-x86_64-w64-mingw32-1.0.9/sourcecpp_69f04dc58d9

C:/PROGRA~1/R/R-42~1.0/bin/x64/R CMD SHLIB -d --preclean -o "sourceCpp_19.dll" "sfgkanal.cpp" 
g++ -std=gnu++11  -I"C:/PROGRA~1/R/R-42~1.0/include" -DNDEBUG   -I"C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include" -I"H:/jupyter-mktmicro-research/R/sfgk"   -I"C:/rtools42/x86_64-w64-mingw32.static.posix/include"     -O2 -Wall -gdwarf-2 -mfpmath=sse -msse2 -mstackrealign  -c sfgkanal.cpp -o sfgkanal.o
In file included from C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include/Rcpp/Vector.h:54,
                 from C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include/Rcpp.h:40,
                 from sfgkanal.cpp:1:
C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include/Rcpp/vector/traits.h: In lambda function:
C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include/Rcpp/vector/traits.h:49:47: warning: '*((void*)(&<anonymous>)+16).Rcpp::traits::r_vector_cache<13, Rcpp::PreserveStorage>::start' may be used uninitialized in this function [-Wmaybe-uninitialized]
   49 |   inline proxy ref(R_xlen_t i) const { return start[i] ; }
      |                                               ^~~~~
C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include/Rcpp/vector/traits.h:49:47: warning: '*((void*)(&<anonymous>)+16).Rcpp::traits::r_vector_cache<13, Rcpp::PreserveStorage>::start' may be used uninitialized in this function [-Wmaybe-uninitialized]
   49 |   inline proxy ref(R_xlen_t i) const { return start[i] ; }
      |                                               ^~~~~
C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include/Rcpp/vector/traits.h:49:47: warning: '*((void*)(&<anonymous>)+16).Rcpp::traits::r_vector_cache<13, Rcpp::PreserveStorage>::start' may be used uninitialized in this function [-Wmaybe-uninitialized]
   49 |   inline proxy ref(R_xlen_t i) const { return start[i] ; }
      |                                               ^~~~~
C:/Users/.../AppData/Local/R/win-library/4.2/Rcpp/include/Rcpp/vector/traits.h:49:47: warning: '*((void*)(&<anonymous>)+16).Rcpp::traits::r_vector_cache<13, Rcpp::PreserveStorage>::start' may be used uninitialized in this function [-Wmaybe-uninitialized]
   49 |   inline proxy ref(R_xlen_t i) const { return start[i] ; }
      |                                               ^~~~~
g++ -std=gnu++11 -shared -static-libgcc -o sourceCpp_19.dll tmp.def sfgkanal.o -LC:/rtools42/x86_64-w64-mingw32.static.posix/lib/x64 -LC:/rtools42/x86_64-w64-mingw32.static.posix/lib -LC:/PROGRA~1/R/R-42~1.0/bin/x64 -lR

问题解答 #

1. windowsDebugDLL=TRUE无效，仍生成优化代码 #

从编译输出可见，g++参数包含-O2（二级优化）和-DNDEBUG（禁用调试宏），说明windowsDebugDLL=TRUE未正确覆盖默认编译参数。解决方法：

• 强制覆盖编译参数，在sourceCpp调用中添加cppflags="-O0 -UNDEBUG -gdwarf-2"，并设置rebuild=TRUE（避免复用缓存）：

  Rcpp::sourceCpp("sfgkanal.cpp", verbose=TRUE, rebuild=TRUE, cacheDir = getwd(), 
                  windowsDebugDLL=TRUE, cppflags="-O0 -UNDEBUG -gdwarf-2")
  

2. -lR链接的是哪个库？ #

在Windows MinGW环境下，-lR实际链接的是R.dll，位于C:/PROGRA~1/R/R-42~1.0/bin/x64/。MinGW链接器会自动在-L指定的路径中查找lib<name>.a或<name>.dll，R在Windows下仅提供动态链接库R.dll，无需对应的静态库文件。

3. R编译用的Makefile位置？ #

使用R CMD SHLIB时，会在编译目录（即输出中的DIR路径：H:/jupyter-mktmicro-research/R/sfgk/sourceCpp-x86_64-w64-mingw32-1.0.9/sourcecpp_69f04dc58d9/）临时生成Makefile，编译完成后不会自动删除，可直接前往该目录查看。

4. Windows下Makevars文件位置及Debug DLL生成 #

Windows下Makevars.win的正确位置：

• 用户级：C:/Users/<你的用户名>/Documents/.R/Makevars.win（注意是Documents文件夹下的.R目录，不是用户根目录的.R）
• 项目级：直接放在代码目录（H:/jupyter-mktmicro-research/R/sfgk/）下，命名为Makevars.win

要生成Debug DLL，可在Makevars.win中添加以下内容，再重新调用sourceCpp并设置rebuild=TRUE：

CXXFLAGS += -O0 -UNDEBUG -gdwarf-2

确认Rtools版本与R版本匹配（你当前使用R-4.2.0对应Rtools42，路径已正确配置），即可生成Debug版本DLL。

内容的提问来源于stack exchange，提问作者Ramesh Kadambi