kalibera/cran-checks

cran-checks's Issues

rsparse package

Hi Tomas. Thanks for all your great work.
I'm trying to debug issue with my rsparse package and as suggested i'm running:

export R_CHECK_CONSTANTS=5
R CMD check rsparse_0.3.3.2.tar.gz

However I'm getting:

• using log directory ‘/home/dselivanov/projects/rsparse/rsparse.Rcheck’
• using R version 3.6.0 (2019-04-26)
• using platform: x86_64-pc-linux-gnu (64-bit)
• using session charset: UTF-8
  Error in readBin(con, "raw", n = 512L) :
  cannot get ALTVEC DATAPTR during GC
  Execution halted

Any ideas on what I'm doing wrong (with R_CHECK_CONSTANTS=1 it works fine)?

Thanks in advance.

Some results missing

E.g. https://github.com/kalibera/cran-checks/blob/master/rchk/results/av.out is still at 0.6.0 (updated 3 months ago on cran)

rcnst error in dependencies

CRAN checks list a rcnst error for the quantities package, but the rcnst report shows that the error happens in dplyr's namespace. I believe that these errors should be attributed to the offending namespace, regardless of the package being checked.

The curl package

I am trying to understand some warnings in the curl package. I hope that discussing this here might benefit others as well. This version concerns curl_2.7. The first warning is about handle.c

Function R_handle_setopt
  [UP] unprotected variable optnames while calling allocating function Rf_asInteger curl/src/handle.c:203
  [UP] unprotected variable optnames while calling allocating function Rf_asReal curl/src/handle.c:223

The only place where R_handle_setopt is invoked is via .Call() and therefore SEXP values should automatically be protected. As I undertand it, this also protects all the elements and attributes of values. So why does this need a protect? Does getAttrib() make a copy??

SEXP optnames = getAttrib(values, R_NamesSymbol);

In addition I don't understand why Rf_asInteger() is called an allocation function? I assumed this to be a readonly function, shorthand for INTEGER(x)[0].

Canonical way to solve spurious protection rchk error with destructors?

I think I have a classical issue of a spurious warning as mentioned in the official docs . How would I go about silencing the rchk CRAN message?

My rchk warning looks like this:

Function Unnester::process(SEXPREC*, SEXPREC*)
  [UP] unprotected variable <unnamed var:   %4 = alloca %struct.SEXPREC*, align 8> while calling allocating function NodeAccumulator::~NodeAccumulator() unnest/src/./unnest.h:284

The error comes from this block, and while I understand the reasoning I don't quite see how to fix it elegantly.

What I came up with is this workaround:

      SEXP out;
      {
        NodeAccumulator acc;
        add_node(*this, acc, vacc, NilSpec, spec, 0, x);
        out = PROTECT(build_df(acc, vacc));
      }
      UNPROTECT(1);
      return out;

Is this a recommended way to silent the warning? If so, it might be useful to add it to the official doc page. For guys like me the solution is not immediately apparent.

Thanks!