[R] robustbase adjbox segfault - memory not mapped

Martin Maechler maechler at stat.math.ethz.ch
Wed Mar 6 09:45:16 CET 2013


>>>>> "B" == Baan  <baanbapat at gmail.com>
>>>>>     on Mon, 4 Mar 2013 22:47:10 +0530 writes:

    B> Thank you Martin. Look forward to the fix.

Committed to the R-forge version of robustbase.

It was a simple integer overflow, indeed, 
"necessarily" happening when the sample size was >= 2^16.5.

I'm planning to submit  robustbase_0.9-7  to CRAN today.
Martin

    B> Regards
    B> Baan


    B> On Monday 04 March 2013 10:19 PM, Martin Maechler wrote:
    >>>>>>> "B" == Baan  <baanbapat at gmail.com>
    >>>>>>> on Mon, 4 Mar 2013 15:02:02 +0530 writes:
    B> Hi, I encountered a segfault, memory not mapped error
    B> when using adjbox in robustbase. In trying to recreate
    B> the issue I found that the error occurs only for large
    B> sample size. Here is the code.
    >> 
    >> >> require(robustbase)
    B> Loading required package: robustbase
    >> >> x <- rnorm(10)
    >> >> y <- rep(1, 10)
    >> >> adjbox(x ~ y) ## gives a plot
    >> >> x <- rnorm(10000)
    >> >> y <- rep(1, 10000)
    >> >> adjbox(x ~ y) ## gives a plot
    >> >> x <- rnorm(100000)
    >> >> y <- rep(1, 100000)
    >> >> adjbox(x ~ y)
    >> 
    B> *** caught segfault ***
    B> address 0xfffffffcc47af530, cause 'memory not mapped'
    >> 
    >> 
    B> Traceback:
    B> 1: .C(mc_C, x, n, eps = eps, iter = c.iter, medc = double(1))
    B> 2: mcComp(x, doReflect, eps1 = eps1, eps2 = eps2, maxit = maxit,
    B> trace.lev = trace.lev)
    B> 3: mc.default(x, ..., na.rm = TRUE)
    B> 4: mc(x, ..., na.rm = TRUE)
    B> 5: adjboxStats(unclass(groups[[i]]), coef = range, doReflect = doReflect)
    B> 6: adjbox.default(split(mf[[response]], mf[-response]), ...)
    B> 7: adjbox(split(mf[[response]], mf[-response]), ...)
    B> 8: adjbox.formula(x ~ y)
    B> 9: adjbox(x ~ y)
    >> 
    >> Indeed, I (as maintainer of robustbase) can reproduce the
    >> segfault.... *even* though you did not specify the random seed...
    >> 
    >> So this should be fixed ... hopefully within a week or so,
    >> but I am not promising anything, given my busy schedule!
    >> 
    >> Martin Maechler,
    >> ETH Zurich
    >> 
    >> [................]
    >> 
    B> My setup details:
    >> 
    B> R --version
    B> R version 2.15.2 (2012-10-26) -- "Trick or Treat"
    >> 
    B> Package:            robustbase
    B> Version:            0.9-5
    B> Date:               2012-03-01
    B> Packaged:           2013-03-01 16:34:03 UTC; maechler
    B> NeedsCompilation:   yes
    B> Repository:         CRAN
    B> Date/Publication:   2013-03-01 18:31:33
    B> Built:              R 2.15.2; x86_64-pc-linux-gnu; 2013-03-04 05:54:20
    B> UTC; unix
    >> 
    >> 
    B> Platform: x86_64-pc-linux-gnu (64-bit)
    B> uname -a
    B> Linux R 2.6.32-5-amd64 #1 SMP Mon Feb 25 00:26:11 UTC 2013 x86_64 GNU/Linux
    B> Debian squeeze
    >> 
    B> Could someone pls help.
    >> 
    B> Regards
    B> Baan



More information about the R-help mailing list