Invalid R code causes invalid registration interface, and cryptic error on compilation

[richfitz]

If the package that is the target of cpp_register contains invalid R code, then the failure of tools::package_native_routine_registration_skeleton leaves the package in an unloadable state with no indication of the underlying problem:

Reprex:

path <- tempfile()
dir.create(path)
dir.create(file.path(path, "src"))
dir.create(file.path(path, "R"))
writeLines(c("Package: pkg", "Version: 0.1.0", "LinkingTo: cpp11"),
           file.path(path, "DESCRIPTION"))
writeLines("useDynLib(pkg, .registration = TRUE)",
           file.path(path, "NAMESPACE"))
writeLines(c("[[cpp11::register]]",
             "double add(double a, double b) {",
             "  return a + b;",
             "}"),
           file.path(path, "src", "code.cpp"))
## Any syntactically incorrect R:
writeLines("f <- function(x", file.path(path, "R", "code.R"))
cpp11::cpp_register(path)
pkgload::load_all(path)

which gives the error:

Loading pkg
ℹ 1 functions decorated with [[cpp11::register]]
✔ generated file cpp11.R
✔ generated file cpp11.cpp
Re-compiling pkg
─  installing *source* package ‘pkg’ ...
   ** using staged installation
   ** libs
   g++ -std=gnu++11 -I"/usr/share/R/include" -DNDEBUG  -I'/home/rich/lib/R/library/cpp11/include' -g -Wall -Wextra -pedantic -Wmaybe-uninitialized -Wno-unused-parameter -Wno-unused-variable -O2   -fpic  -g -O2 -fdebug-prefix-map=/build/r-base-7BvS0x/r-base-4.0.3=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -g  -c code.cpp -o code.o
   code.cpp:2:30: warning: ‘cpp11::register’ scoped attribute directive ignored [-Wattributes]
    double add(double a, double b) {
                                 ^
   g++ -std=gnu++11 -I"/usr/share/R/include" -DNDEBUG  -I'/home/rich/lib/R/library/cpp11/include' -g -Wall -Wextra -pedantic -Wmaybe-uninitialized -Wno-unused-parameter -Wno-unused-variable -O2   -fpic  -g -O2 -fdebug-prefix-map=/build/r-base-7BvS0x/r-base-4.0.3=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -g  -c cpp11.cpp -o cpp11.o
   cpp11.cpp: In function ‘void R_init_pkg(DllInfo*)’:
   cpp11.cpp:20:33: error: ‘CallEntries’ was not declared in this scope
      R_registerRoutines(dll, NULL, CallEntries, NULL, NULL);
                                    ^~~~~~~~~~~
   /usr/lib/R/etc/Makeconf:181: recipe for target 'cpp11.o' failed
   make: *** [cpp11.o] Error 1
   ERROR: compilation failed for package ‘pkg’
─  removing ‘/tmp/RtmpWQKwSR/devtools_install_6916495ebf34/pkg’
Error in (function (command = NULL, args = character(), error_on_status = TRUE,  : 
  System command 'R' failed, exit status: 1, stdout + stderr (last 10 lines):
E>                               ^
E> g++ -std=gnu++11 -I"/usr/share/R/include" -DNDEBUG  -I'/home/rich/lib/R/library/cpp11/include' -g -Wall -Wextra -pedantic -Wmaybe-uninitialized -Wno-unused-parameter -Wno-unused-variable -O2   -fpic  -g -O2 -fdebug-prefix-map=/build/r-base-7BvS0x/r-base-4.0.3=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -g  -c cpp11.cpp -o cpp11.o
E> cpp11.cpp: In function ‘void R_init_pkg(DllInfo*)’:
E> cpp11.cpp:20:33: error: ‘CallEntries’ was not declared in this scope
E>    R_registerRoutines(dll, NULL, CallEntries, NULL, NULL);
E>                                  ^~~~~~~~~~~
E> /usr/lib/R/etc/Makeconf:181: recipe for target 'cpp11.o' failed
E> make: *** [cpp11.o] Error 1
E> ERROR: compilation failed for package ‘pkg’
E> * removing ‘/tmp/RtmpWQKwSR/devtools_insta
[...]
Type .Last.error.trace to see where the

The issue seems to stem fro get_call_entries wrapping tools::package_native_routine_registration_skeleton in a silent try:

  try(
    tools::package_native_routine_registration_skeleton(path,
      con,
      character_only = FALSE,
      include_declarations = TRUE
    ),
    silent = TRUE
  )

without which the underlying error would have been revealed:

Error in parse(file, keep.source = keep.source) : 
  /tmp/RtmpAK7co3/file699a2cbec253/R/code.R:2:0: unexpected end of input
1: f <- function(x
   ^

We've seen this a couple of times in the wild and it's a head-scratcher each time.

[jimhester]

I have ran into this as well, perhaps we can just remove the try.