I don't know much about the iMac. R's .C() passes
R-language integers as pointers to 32-bit ints. If
your iMac is 64-bit and sizeof(long)==8 in your compiler
(pretty common for 64-bit compilers but not so for
Microsoft Windows compilers) then Long[0] will
use the next 64 bits to make an integral value out of
while Int[0] will only use the next 32 bits. Since R
only passes 32 bits/integer your Long[0] would include
32 random bits in its value and using it as a subscript
would often point to a place in memory that you don't own.
Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com
________________________________
From: David [mailto:david.mailli...@gmail.com]
Sent: Wednesday, October 13, 2010 11:13 AM
To: William Dunlap
Cc: r-help@r-project.org
Subject: Re: [R] "Memory not mapped" when using .C,problem in
Mac but not in Linux
Thank you very much for your kind reply.
I used gdb, and it returns a reason "KERN_INVALID_ADDRESS" on a
very simple operation (obtaining the time step for a numerical
integration). Please see bellow.
But, um, I solved it by changing the function's arguments in the
C code from "unsigned long int" to "int". In R, those arguments were
defined with "as.integer". So, can you now understand why the error
happens (on iMac and not on Linux)? If yes, can you explain it to me?
:-)
If some argument is defined in C as "unsigned long int", I see
that the R equivalent is *not* "as.integer". Which is it?
Thank you very much in advance,
David
Can you please Cc to me any replies, just in case I may miss any
of them among the whole amount of emails :-) ?
---------------------------------------
Program received signal EXC_BAD_ACCESS, Could not access memory.
Reason: KERN_INVALID_ADDRESS at address: 0x0000000902ef3d68
[...]
139 double
time_step=(times[number_times[0]-1]-times[0])/(number_timesteps[0]);
(note: "number_times" and "number_timesteps" are integer arrays
with a single component, because they are arguments of the C function
that need to be passed from R. They were defined as "unsigned long int"
in C, and "as.integer" in R, giving the error. If I define them as "int"
in C, the error disappears).
2010/10/13 William Dunlap <wdun...@tibco.com>
This often happens when your C code uses memory that
it did not allocate, particularly when it reads or
writes just a little beyond the end of a memory block.
On some platforms or if you are lucky there is unused
memory between blocks of allocated memory and you don't
see a problem. Other machines may pack allocated memory
blocks more tightly and writing off the end of an array
corrupts another array, leading to the program crashing.
(Or reading off the end of an array may pick up data
from
the next array, which leads to similar crashes later
on.)
You may also be reading memory which has never been
written
to, hence you are using random data, which could corrupt
things.
Use a program like valgrind (on Linux, free) or Purify
(on
various platforms, but pricy) to detect memory misuse
in C/C++ code. As long as your code isn't #ifdef'ed for
various platforms, you can use valgrind to detect and
fix
memory misuse on Linux and you will find that overt
problems on other platforms will go away.
Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com
> -----Original Message-----
> From: r-help-boun...@r-project.org
> [mailto:r-help-boun...@r-project.org] On Behalf Of
David
> Sent: Wednesday, October 13, 2010 6:47 AM
> To: r-help@r-project.org
> Subject: [R] "Memory not mapped" when using .C,problem
in Mac
> but not in Linux
>
> Hello,
>
> I am aware this may be an obscure problem difficult to
advice
> about, but
> just in case... I am calling a C function from R on an
iMac
> (almost shining:
> bought by my institution this year) and gives a
"memory not
> mapped" error.
>
> Nevertheless, exactly the same code runs without
problem in a powerful
> computer using SuSE Linux, but also on my laptop of
2007, 32
> bits, 2 GB RAM,
> running Ubuntu. My supervisor says that he can run my
code on
> his iMac (a
> bit older than mine) without problem.
>
> I have upgraded to the latest version of R, and I have
tried
> compiling the C
> code with the latest version of gcc (obtained from
MacPorts),
> but the error
> persists.
>
> Any ideas?
>
> Thank you very much in advance,
>
> David
> Can you please Cc to me any replies, just in case I
may miss
> any of them
> among the whole amount of emails :-) ?
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> 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.
>
[[alternative HTML version deleted]]
______________________________________________
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.
