R devel - Mar 2004 - qnorm(2) ends with segmentation fault (PR#6648)

Full_Name: Xiong Guanglei
Version: 1.8.1
OS: Linux
Submission from: (NULL) (202.38.103.50)


qnorm(x)
when x>1.0

linuxfavor@sohu.com wrote:
> Full_Name: Xiong Guanglei
> Version: 1.8.1
> OS: Linux
> Submission from: (NULL) (202.38.103.50)
> 
> 
> qnorm(x)
> when x>1.0
is not defined, hence Not a Number: NaN
Obviously not a bug.

Uwe Ligges

> ______________________________________________
> R-devel@stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-devel

Which Linux is that, and is R compiled from source?  I tested both on SLES8
for amd64 and Mandrake 9 on Xeon, compiled R from source in both.  I just
get NaN and a warning about the NaN.  (I tried qnorm(2) as the subject
says.)

Andy
> From: linuxfavor@sohu.com
> 
> 
> Full_Name: Xiong Guanglei
> Version: 1.8.1
> OS: Linux
> Submission from: (NULL) (202.38.103.50)
> 
> 
> qnorm(x)
> when x>1.0
> 

------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachments,...{{dropped}}

Which Linux is that, and is R compiled from source?  I tested both on SLES8
for amd64 and Mandrake 9 on Xeon, compiled R from source in both.  I just
get NaN and a warning about the NaN.  (I tried qnorm(2) as the subject
says.)

Andy
> From: linuxfavor@sohu.com
> 
> 
> Full_Name: Xiong Guanglei
> Version: 1.8.1
> OS: Linux
> Submission from: (NULL) (202.38.103.50)
> 
> 
> qnorm(x)
> when x>1.0
> 

------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachments,...{{dropped}}

>>>>> "XiGu" == linuxfavor  <linuxfavor@sohu.com>
>>>>>     on Fri,  5 Mar 2004 15:21:15 +0100 (CET) writes:
    XiGu> Full_Name: Xiong Guanglei Version: 1.8.1 OS: Linux
    XiGu> Submission from: (NULL) (202.38.103.50)


    XiGu> qnorm(x) when x>1.0

produces NaN (and a warning) for me --- as it should --- 
both in R 1.8.1 and  R-devel.

Can you give a bit more details about how you got your version
of R (binary / compiled yourself, ...)?

Do other people see seg.faults for
  qnorm(2) 
?

Regards,
Martin

On Fri, 5 Mar 2004 ligges@statistik.uni-dortmund.de wrote:
> linuxfavor@sohu.com wrote:
> 
> > Full_Name: Xiong Guanglei
> > Version: 1.8.1
> > OS: Linux
> > Submission from: (NULL) (202.38.103.50)
> > 
> > 
> > qnorm(x)
> > when x>1.0
> 
> is not defined, hence Not a Number: NaN
> Obviously not a bug.
It is a bug if he gets a segfault, but I do not (and I suppose neither
does Uwe).  As the value is checked in src/nmath/dnorm.c, a segfault seems
to indicate a badly broken build of R.

So, what is meant by `Linux', and precisely what compilers, kernel and 
libc were used?

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

On Fri, 2004-03-05 at 09:34, ligges@statistik.uni-dortmund.de
wrote:
> linuxfavor@sohu.com wrote:
> 
> > Full_Name: Xiong Guanglei
> > Version: 1.8.1
> > OS: Linux
> > Submission from: (NULL) (202.38.103.50)
> > 
> > 
> > qnorm(x)
> > when x>1.0
> 
> is not defined, hence Not a Number: NaN
> Obviously not a bug.

Right, but it should not segfault, which it does not in Version 1.8.1
Patched (2004-03-04), so presumably this error has been trapped.
> qnorm(2)
[1] NaN
Warning message: 
NaNs produced in: qnorm(p, mean, sd, lower.tail, log.p) 

HTH,

Marc Schwartz