Re: [R] R crashes when loading rgl package before minqa package

[Gaspard Lequeux]

Hej,

On Thu, 30 Sep 2010, Ravi Varadhan wrote:

I get this on Windows (it does not crash):

library(minqa)
library(rgl)
newuoa(initpar, optimft)

Error in newuoa(initpar, optimft) :
 non-finite x values not allowed in calfun
In addition: Warning message:
In log(x[4]) : NaNs produced

Does it crash when you load first rgl and then only minqa? Like this:

library(rgl)
library(minqa)
newuoa(initpar, optimft)

/Gaspard

This tells me that you should be constraining your parameter x[4] (may be even 
x[5]) to be non-negative:

Here is what I get with `bobyqa':

bobyqa(initpar, optimft, lower=c(-Inf, -Inf, -Inf, 0, 0))

parameter estimates: -5.311767080681, -3861.89005072333, 979.239647766226, 
0.268156271922112, 27.6418856936228
objective: 1457.20987728737
number of function evaluations: 78

Ravi.

____________________________________________________________________

Ravi Varadhan, Ph.D.
Assistant Professor,
Division of Geriatric Medicine and Gerontology
School of Medicine
Johns Hopkins University

Ph. (410) 502-2619
email: rvarad...@jhmi.edu


----- Original Message -----
From: Gaspard Lequeux <gaspard.lequ...@biomath.ugent.be>
Date: Wednesday, September 29, 2010 11:40 am
Subject: [R] R crashes when loading rgl package before minqa package
To: r-help@r-project.org

 Hej,

 Calling newuoa (from the minqa package) makes R crash when the
package rgl is loaded first. This however only on certain selected data.

 The data used for testing (saved to 'bugs.R'):


 xvals = 
c(1,2,4,5,7,8,9,10,11,12,14,15,16,18,19,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36)

 yvals = 
c(857.7597,975.8624,978.2655,979.3034,965.5919,983.8946,992.2512,992.1178,979.5379,974.4269,968.4113,991.5210,977.3361,985.7800,975.5220,974.6880,973.8102,980.7295,982.0034,984.7993,978.4948,970.4351,969.0718,983.7892,976.3637,980.7833,987.1665,976.6000,975.1332,971.0757,989.4693)

 initpar = c(-5.1471384, -3861.8905839, 979.2616002, 0.2572355, 27.5705764)

 optimft <- function(x) {
   yft = x[2] + (x[3] - x[2])/((1 + exp(x[1] * (log(xvals) - log(x[4]))))^x[5])
   return(sum((yvals - yft)^2))
 }


 Sequence of commands needed to make the bug appear:

 <Start R>
 source('bugs.R')
 library(minqa)
 library(rgl)
 newuoa(initpar, optimft)
  => OK

 <Start R>
 source('bugs.R')
 library(rgl)
 library(minqa)
 newuoa(initpar, optimft)
   => Crash: segfault: address 0x18, cause 'memory not mapped'

 I found the bug using the package qpcR, where rgl is loaded when
loading qpcR while minqa is only loaded later, when needed.


 Running on Debian squeeze 64 bit.
 R version: R version 2.11.1 (2010-05-31) x86_64-pc-linux-gnu
 rgl version: 0.91
 minqa version:  1.1.9
 Rcpp version: 0.8.6 (loaded by minqa)

 Kind regards,

 Gaspard Lequeux

 ______________________________________________
 R-help@r-project.org mailing list

 PLEASE do read the posting guide
 and provide commented, minimal, self-contained, reproducible code.

______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.