Skip to content

[R-pkg-devel] clang issue

10 messages · Grose, Daniel, Ivan Krylov, Dirk Eddelbuettel

Original
Grose, Daniel
# 
Hi

My package submission fails due to --------



In file included from MeanAnomalyMV.cpp:7:
In file included from ./Functions_mean.h:4:
In file included from usr/local/clang-trunk/bin../include/c++/v1/vector:321:
In file included from
usr/local/clang-trunk/bin../include/c++/v1/__format/formatter_bool.h:20:
In file included from
usr/local/clang-trunk/bin../include/c++/v1/__format/formatter_integral.h:32:
In file included from usr/local/clang-trunk/bin../include/c++/v1/locale:202:
usr/local/clang-trunk/bin../include/c++/v1/__locale:962:34: error: too
many arguments provided to function-like macro invocation
   962 |     int length(state_type& __st, const extern_type* __frm,
const extern_type* __end, size_t __mx) const
       |                                  ^
/data/gannet/ripley/R/R-flang/include/Rinternals.h:968:9: note: macro
'length' defined here
   968 | #define length(x)               Rf_length(x)
       |         ^
...

Please define R_NO_REMAP and fix the fallout: it has Rinternals.h
included all over the place in both .h and .cpp files.  It must not be
included before <vector> nor any other C++ header.

Please fix and resubmit.

--------------------------------------------

I have tried to reproduce the error using clang-15 on Ubuntu-22.04, but everything seems fine --

 clang-15 -std=gnu++17 -I"/usr/share/R/include" -DNDEBUG -DBOOST_ALLOW_DEPRECATED_HEADERS -DR_NO_REMAP -I'/home/grosedj/R-packages/Rcpp/include' -I'/home/grosedj/R-packages/BH/include'    -fpic  -g -O2 -fdebug-prefix-map=/build/r-base-a3XuZ5/r-base-4.2.2.20221110=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2  -c MeanAnomalyMV.cpp -o MeanAnomalyMV.o


Any ideas ?

Many thanks,

Daniel Grose
Dirk Eddelbuettel
# 
Daniel,

This is not new, and not related to clang.
On 7 August 2023 at 12:58, Grose, Daniel wrote:
| Any ideas ?

Add a line

   #define R_NO_REMAP 1

before _any_ inclusion of R headers. See Section 6 of Writing R Extensions.

If you use eg Rcpp it is done for you when you include Rcpp headers.

Cheers, Dirk
Grose, Daniel
# 
Hi Dirk

Thanks for the quick response.

I thought  -DR_NO_REMAP in Makevars would set R_NO_REMAP across all compilation units. Is this not the case ?

Thanks,

Dan
Ivan Krylov
# 
? Mon, 7 Aug 2023 13:16:33 +0000
"Grose, Daniel" <dan.grose at lancaster.ac.uk> ?????:
It should have worked. Can we see the code? Is there a visible compiler
command line in the check results that you quoted in the first message?
Grose, Daniel
# 
Hi Ivan

No - unfortunately not. I cannot reproduce the error locally so I posted the snippet that Uwe Ligges sent me regarding the failed CRAN submission. That is all of the information I have.

For now I will follow Dirk's advice and see if the CRAN submission gets accepted.

Thanks for your help.

Best wishes,
Dan
Dirk Eddelbuettel
#
On 7 August 2023 at 13:38, Grose, Daniel wrote:
| No - unfortunately not. I cannot reproduce the error locally so I posted the
| snippet that Uwe Ligges sent me regarding the failed CRAN submission. That is
| all of the information I have.
| 
| For now I will follow Dirk's advice and see if the CRAN submission gets
| accepted.

Yes as Ivan state using PKG_CFLAGS etc should be equivalent but there can be
pitfalls if you eg set PKG_CFLAGS but also have a C++ (or Fortran) file.

So as Ivan already asked: what is you repo?  We can help better (and faster)
with reproducible code.

Best, Dirk
Grose, Daniel
# 
Hi Ivan/Dirk

As I say - I cannot reproduce the error myself. I will try asking Uwe Ligges for more information.

Best wishes,
Dan
Dirk Eddelbuettel
#
On 7 August 2023 at 13:55, Grose, Daniel wrote:
| As I say - I cannot reproduce the error myself. I will try asking Uwe Ligges
| for more information.

Methinks you are doing it wrong. We created the r-package-devel list years
ago to take load away from the overworked CRAN maintainers who are "on the
critical path" and a bottleneck we would rather see lightened.

Also, you were asked twice to share your repo or package. As you won't I
stop here as you seemingly do not want us to help you. Your call.

Dirk
Grose, Daniel
# 
Hi Dirk

My apologies. I thought you were asking for compiler output. You can find the code at


https://github.com/grosed/anomaly

[https://opengraph.githubassets.com/866d37f560339c5ee145d5cd7e40edf768223cf08b8e9f35ae97e14d1f2b47f5/grosed/anomaly]<https://github.com/grosed/anomaly>
GitHub - grosed/anomaly<https://github.com/grosed/anomaly>
Contribute to grosed/anomaly development by creating an account on GitHub.
github.com


in the version-4.3.0 branch.

Many thanks,
Dan
Ivan Krylov
# 
? Mon, 7 Aug 2023 15:52:38 +0000
"Grose, Daniel" <dan.grose at lancaster.ac.uk> ?????:
Ah, so this is why I couldn't find your package on CRAN. Now I can see
the compiler log, together with the compiler command line:
https://www.stats.ox.ac.uk/pub/bdr/clang17/anomaly.log

It looks like -DR_NO_REMAP is missing from src/Makevars, both in the
last archived version on CRAN and in the version-4.3.0 branch, but if
you add it to the line with PKG_CPPFLAGS, it should work.

If it's too much to fix, at least make sure to move any R-related
#includes below the system #includes. MeanAnomalyMV.cpp originally blew
up because (1) -DR_NO_REMAP was missing from the compiler command line,
(2) Rinternals.h was included, which provides #define length(x), and
(3) a standard C++ header was included, which tried to define a
function named length. Moving the R headers below the standard C++
headers would have prevented it.

It may also help to change the interface of MeanAnomalyMV so that it
either (1) uses Rcpp, enjoying built-in R_NO_REMAP or (2) works purely
in terms of R and C types, avoiding the need to #include <vector> and
<string>.

Hope you'll succeed in getting the package back in shape!