Segfault on `stop`

[romainfrancois]

This has been reported by @pschmied here initially. tidyverse/dplyr#1318

maui% cat ./test.cpp 
#include <Rcpp.h>
using namespace Rcpp;

// [[Rcpp::export]]
void test(){
  SEXP s = Rf_install("County") ;
  CharacterVector v(1) ; v[0] = PRINTNAME(s) ;
  const char* txt = v[0];
  stop( "unknown column '%s' ", txt ) ;
}


/*** R
  test()
*/
maui% Rscript -e "Rcpp::sourceCpp( '~/tmp/test.cpp') "

> test()
Error: segfault from C stack overflow
Error: C stack usage  121890565849832 is too close to the limit
Execution halted
maui% 

I can't reproduce it locally. Perhaps someone with a system closer to @pschmied. I guess it's an issue with tinyformat.

[eddelbuettel]

I cannot (thanks, Romain, now edited) reproduce it with R 3.2.2 and Rcpp 0.12.0.2:

edd@max:~$ cat /tmp/issue347.cpp    
#include <Rcpp.h>
using namespace Rcpp;

// [[Rcpp::export]]
void test(){
  SEXP s = Rf_install("County") ;
  CharacterVector v(1) ; v[0] = PRINTNAME(s) ;
  const char* txt = v[0];
  stop( "unknown column '%s' ", txt ) ;
}

/*** R
test()
cat("Done\n")
*/
edd@max:~$ Rscript -e 'Rcpp::sourceCpp("/tmp/issue347.cpp")'  

R> test()
Error in eval(expr, envir, enclos) : unknown column 'County' 
Calls: <Anonymous> ... source -> withVisible -> eval -> eval -> test -> <Anonymous>
Execution halted
edd@max:~$ 

The output is a little cleaner if we use r from littler:

edd@max:~/git/rcpp/inst/unitTests(master)$ r -e 'Rcpp::sourceCpp("/tmp/issue347.cpp")'

> test()
Error in eval(expr, envir, enclos) : unknown column 'County' 
edd@max:~/git/rcpp/inst/unitTests(master)$ 

[romainfrancois]

@eddelbuettel what you have is the expected behavior. I guess you meant "can't".

[eddelbuettel]

Ahh yes. Early in the day and no coffee yet (but on the third bug report ...). Now edited. Thanks.

Well if that is the expected behaviour, what would the error be? What is his platform? We'd still need something reproducible. If your (OS X, newer clang) and my (Linux, g++ 4.9.2) systems do not show, maybe the bug is dependent on an old system?

[romainfrancois]

@pschmied gets a segfault: not intended.
Otherwise a call to stop generating an error that aborts an Rscript: definitely what is intended. You could wrap it in a try to see "done"

[eddelbuettel]

Well but without a minimally reproducible example there is little for us to work on.

Maybe you feel motivated to work on this, it seems to itch you a little harder than me.

[romainfrancois]

Hmm i don't really care. I just passed on the information so that i could close the issue on the dplyr repo as this was not a dplyr problem.

[eddelbuettel]

Not a Rcpp issue either as there is no reproducible example.

[pschmied]

Hi there, sorry I'm late to the party. And thank you, Romain, for kicking this over here. I'd intended to do that, and it slipped down my to-do list.

The error is reproducible on OpenBSD (at least the most recent release and snapshot versions, which are what I've tested on). I need to rebuild Rcpp with debugging symbols and then put through gdb. Any chance you'd be willing to reopen this / work with me on figuring out the issue?

If the issue is ultimately with OpenBSD, I'm sure they'll be glad for the history, and I can bump this to them.

[pschmied]

Here is the debugger output. Can someone more versed in reading GDB backtraces confirm that the problem looks to originate upstream of Rcpp on this?

maui% R -d gdb -e "Rcpp::sourceCpp( '~/tmp/test.cpp') "       
GNU gdb 6.3
Copyright 2004 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "amd64-unknown-openbsd5.8"...(no debugging symbols found)

(gdb) c
The program is not being run.
(gdb) r
Starting program: /usr/local/lib/R/bin/exec/R -e Rcpp::sourceCpp\(\~+\~\'\~/tmp/test.cpp\'\)\~+\~

R version 3.2.1 (2015-06-18) -- "World-Famous Astronaut"
Copyright (C) 2015 The R Foundation for Statistical Computing
Platform: x86_64-unknown-openbsd5.8 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> Rcpp::sourceCpp( '~/tmp/test.cpp') 

> test()

Program received signal SIGSEGV, Segmentation fault.
0x00007f7ffffc0df8 in ?? ()
(gdb) bt
#0  0x00007f7ffffc0df8 in ?? ()
#1  0x0000084ca004c628 in tinyformat::detail::formatImpl (out=@0x7f7ffffc0da0, fmt=0x84ca014d747 "%s' ", formatters=0x7f7ffffc0d40, numFormatters=1) at tinyformat.h:495
#2  0x0000084ca004d1ea in tinyformat::format<char const*> (fmt=0x84ca014d737 "unknown column '%s' ", v1=@0x7f7ffffc0f88) at tinyformat.h:1018
#3  0x0000084ca0049eca in test () at exceptions.h:256
#4  0x0000084ca004a095 in sourceCpp_0_test () at test.cpp:27
#5  0x0000084ce58f10a6 in do_dotcall () from /usr/local/lib/R/lib/libR.so.3.1
#6  0x0000084ce592c9f7 in Rf_eval () from /usr/local/lib/R/lib/libR.so.3.1
#7  0x0000084ce592ff8a in Rf_evalList () from /usr/local/lib/R/lib/libR.so.3.1
#8  0x0000084ce592c8c6 in Rf_eval () from /usr/local/lib/R/lib/libR.so.3.1
#9  0x00007f7ffffc1e10 in ?? ()
#10 0x000000000000001e in ?? ()
#11 0x0000000000000000 in ?? ()
(gdb) 

[eddelbuettel]

Hi @pschmied -- of course I'd be happy to work with you for as long as you can play along and test changes. I am sure we can nail this down somewhere between tinyformat and *BSD. I don't suspect this to be particular to anyone of us (ie dplyr, Rcpp, R, ...) -- it seems like an invalid string is propagated.

Can you lead this from your end? Maybe get Rcpp to print the format string etc in detail and then use that in a smaller standalone command-line program against tinyformat?

[pschmied]

Thanks, @eddelbuettel! I'll give that a go. My C/C++ chops aren't quite where I'm at with R, but I'm fearless and armed with righteous ignorance ;-)

[eddelbuettel]

No worries @pschmied that combination has served me well for years too :)

[kevinushey]

@pschmied, what version of gcc are you using to compile? (What's the output of gcc -v)?