R help - Jun 2004 - [Q] GET_DIM() crash on Windows only

I have the following contrived code in package format.
On Solaris and Mac OS X, code runs just fine. On Windows,
it crashes the R environment with the "Send Bug Report"
dialog. I tried R 1.8.1 (Win2K) and R 1.9 (WinXP) binaries
with the same result. PCs otherwise appear properly
configured for creating R packages. Anything blatantly
wrong? Suggestions?

TIA



Relevant files from package (getdim):


R/getdim.R
----------
getdim <- function(x)
    .Call("getdim",
          as.matrix(x),
          PACKAGE="getdim");


R/zzz.R
-------
.First.lib <- function(libname, pkgname) {
    if (version$major == 1 && version$minor < 8.1) {
        stop("Requires R 1.8.1 or later")
    }
    library.dynam("getdim", pkgname, libname)
}


src/getdim.c
------------
#include <R.h>
#include <Rdefines.h>
#include <R_ext/PrtUtil.h>

int GetMatrixDimen(SEXP vntX, int *nrow, int *ncol)
{
    SEXP vntXdim;
    int nX;

#ifdef DEBUG_GETDIM
    REprintf("In GetMatrixDimen()...\n");
#endif

/**** Code crashes here on Windoze ******/
    PROTECT(vntXdim = GET_DIM(vntX));
#ifdef DEBUG_GETDIM
    REprintf("\tgot dimension object\n");
#endif

    nX = GET_LENGTH(vntXdim);
#ifdef DEBUG_GETDIM
    REprintf("\tgot length from dimension object\n");
#endif

    if (nX == 2)
    {
        int *piXdim = INTEGER_POINTER(vntXdim);

        *nrow = piXdim[0];
        *ncol = piXdim[1];
    }
    else
    {
        *nrow = -1;
        *ncol = -1;
    }
    UNPROTECT(1);

    return nX;
}


SEXP getdim(SEXP vntX)
{
    SEXP vntOut;
    int m, n;

#ifdef DEBUG_GETDIM
    REprintf("In getdim(x)...\n");
#endif

    if (GetMatrixDimen(vntX, &m, &n) != 2)
    {
        error("'x' is not a two dimensional matrix");
        /*NOTREACHED*/
    }

#ifdef DEBUG_GETDIM
    REprintf("\tm = %d\n", m);
    REprintf("\tn = %d\n", n);
#endif

    PROTECT(vntOut = NEW_INTEGER(2));
    INTEGER(vntOut)[0] = m;
    INTEGER(vntOut)[1] = n;
    UNPROTECT(1);

    return vntOut;
}


tests/getdim.R
--------------
library(getdim)

test.getdim <- function(input, expected) {
    result <- getdim(input);
    identical(all.equal(result, expected), TRUE);
}

mat <- matrix(1:6, 3, 2)
mat.expected <- c(3, 2)
test.getdim(mat, mat.expected)

vec <- 1:6
vec.expected <- c(6, 1)
test.getdim(vec, vec.expected)

----------------------------------------------------------
SIGSIG -- signature too long (core dumped)

Paul Roebuck wrote:
> I have the following contrived code in package format.
> On Solaris and Mac OS X, code runs just fine. On Windows,
> it crashes the R environment with the "Send Bug Report"
> dialog. I tried R 1.8.1 (Win2K) and R 1.9 (WinXP) binaries
> with the same result. PCs otherwise appear properly
> configured for creating R packages. Anything blatantly
> wrong? Suggestions?

Works for me (R-1.9.1, WinNT4.0), even with gctorture(TRUE).
Did you use the recommended compiler and tools?

Uwe Ligges

> TIA
> 
> 
> 
> Relevant files from package (getdim):
> 
> 
> R/getdim.R
> ----------
> getdim <- function(x)
>     .Call("getdim",
>           as.matrix(x),
>           PACKAGE="getdim");
> 
> 
> R/zzz.R
> -------
> .First.lib <- function(libname, pkgname) {
>     if (version$major == 1 && version$minor < 8.1) {
>         stop("Requires R 1.8.1 or later")
>     }
>     library.dynam("getdim", pkgname, libname)
> }
> 
> 
> src/getdim.c
> ------------
> #include <R.h>
> #include <Rdefines.h>
> #include <R_ext/PrtUtil.h>
> 
> int GetMatrixDimen(SEXP vntX, int *nrow, int *ncol)
> {
>     SEXP vntXdim;
>     int nX;
> 
> #ifdef DEBUG_GETDIM
>     REprintf("In GetMatrixDimen()...\n");
> #endif
> 
> /**** Code crashes here on Windoze ******/
>     PROTECT(vntXdim = GET_DIM(vntX));
> #ifdef DEBUG_GETDIM
>     REprintf("\tgot dimension object\n");
> #endif
> 
>     nX = GET_LENGTH(vntXdim);
> #ifdef DEBUG_GETDIM
>     REprintf("\tgot length from dimension object\n");
> #endif
> 
>     if (nX == 2)
>     {
>         int *piXdim = INTEGER_POINTER(vntXdim);
> 
>         *nrow = piXdim[0];
>         *ncol = piXdim[1];
>     }
>     else
>     {
>         *nrow = -1;
>         *ncol = -1;
>     }
>     UNPROTECT(1);
> 
>     return nX;
> }
> 
> 
> SEXP getdim(SEXP vntX)
> {
>     SEXP vntOut;
>     int m, n;
> 
> #ifdef DEBUG_GETDIM
>     REprintf("In getdim(x)...\n");
> #endif
> 
>     if (GetMatrixDimen(vntX, &m, &n) != 2)
>     {
>         error("'x' is not a two dimensional matrix");
>         /*NOTREACHED*/
>     }
> 
> #ifdef DEBUG_GETDIM
>     REprintf("\tm = %d\n", m);
>     REprintf("\tn = %d\n", n);
> #endif
> 
>     PROTECT(vntOut = NEW_INTEGER(2));
>     INTEGER(vntOut)[0] = m;
>     INTEGER(vntOut)[1] = n;
>     UNPROTECT(1);
> 
>     return vntOut;
> }
> 
> 
> tests/getdim.R
> --------------
> library(getdim)
> 
> test.getdim <- function(input, expected) {
>     result <- getdim(input);
>     identical(all.equal(result, expected), TRUE);
> }
> 
> mat <- matrix(1:6, 3, 2)
> mat.expected <- c(3, 2)
> test.getdim(mat, mat.expected)
> 
> vec <- 1:6
> vec.expected <- c(6, 1)
> test.getdim(vec, vec.expected)
> 
> ----------------------------------------------------------
> SIGSIG -- signature too long (core dumped)
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html

Uwe Ligges wrote:
> Paul Roebuck wrote:
> 
>> I have the following contrived code in package format.
>> On Solaris and Mac OS X, code runs just fine. On Windows,
>> it crashes the R environment with the "Send Bug Report"
>> dialog. I tried R 1.8.1 (Win2K) and R 1.9 (WinXP) binaries
>> with the same result. PCs otherwise appear properly
>> configured for creating R packages. Anything blatantly
>> wrong? Suggestions?
> 
> 
> 
> Works for me (R-1.9.1, WinNT4.0), even with gctorture(TRUE).
> Did you use the recommended compiler and tools?
> 
> Uwe Ligges
> 
> 
And for me on R-1.9.0, Win2000.

--sundar

Uwe Ligges wrote:
> Paul Roebuck wrote:
> 
>> I have the following contrived code in package format.
>> On Solaris and Mac OS X, code runs just fine. On Windows,
>> it crashes the R environment with the "Send Bug Report"
>> dialog. I tried R 1.8.1 (Win2K) and R 1.9 (WinXP) binaries
>> with the same result. PCs otherwise appear properly
>> configured for creating R packages. Anything blatantly
>> wrong? Suggestions?
> 
> 
> 
> Works for me (R-1.9.1, WinNT4.0), even with gctorture(TRUE).
> Did you use the recommended compiler and tools?
> 
> Uwe Ligges
> 
> 
And for me on R-1.9.0, Win2000.

--sundar

<snip>