Prev: Matrix to Column vector
Next: Use of uigetfile() together with uiwait()
From: Taylor on 18 Jun 2010 04:50
I have written a mex file that compiles fine within 32-bit windows xp and works without any problems. However, when I tried compiling it under a 64-bit linux environment, I get the following error after compilation (which occurs just fine):

Mex file entry point is missing. Please check the (case-sensitive)
spelling of mexFunction (for C MEX-files), or the (case-insensitive)
spelling of MEXFUNCTION (for FORTRAN MEX-files).
??? Invalid MEX-file

I've searched for why this happens, but I can't seem to find anything useful. My project is in C++ and I'm compiling it with g++. In Windows, if I compile it with visual studio, it works just fine. I have the entry point spelled correctly with the correct arguments. I'm at a loss to why this is happening. Thanks for any help.
From: Steve Amphlett on 18 Jun 2010 05:59
"Taylor " <twsouthwick(a)byu.net> wrote in message <hvfc0g$grj$1(a)fred.mathworks.com>...
> I have written a mex file that compiles fine within 32-bit windows xp and works without any problems. However, when I tried compiling it under a 64-bit linux environment, I get the following error after compilation (which occurs just fine):
>
> Mex file entry point is missing. Please check the (case-sensitive)
> spelling of mexFunction (for C MEX-files), or the (case-insensitive)
> spelling of MEXFUNCTION (for FORTRAN MEX-files).
> ??? Invalid MEX-file
>
> I've searched for why this happens, but I can't seem to find anything useful. My project is in C++ and I'm compiling it with g++. In Windows, if I compile it with visual studio, it works just fine. I have the entry point spelled correctly with the correct arguments. I'm at a loss to why this is happening. Thanks for any help.

First place to start would be a symbol dump from the mexa64 file:

% nm whatever.mexa64

to find out what symbols it actually does have inside it.
From: Rune Allnor on 18 Jun 2010 06:04
On 18 Jun, 10:50, "Taylor " <twsouthw...(a)byu.net> wrote:
> I have written a mex file that compiles fine within 32-bit windows xp and works without any problems.  However, when I tried compiling it under a 64-bit linux environment, I get the following error after compilation (which occurs just fine):
>
> Mex file entry point is missing.  Please check the (case-sensitive)
> spelling of mexFunction (for C MEX-files), or the (case-insensitive)
> spelling of MEXFUNCTION (for FORTRAN MEX-files).
> ??? Invalid MEX-file
>
> I've searched for why this happens, but I can't seem to find anything useful.  My project is in C++ and I'm compiling it with g++.  In Windows, if I compile it with visual studio, it works just fine.  I have the entry point spelled correctly with the correct arguments.  I'm at a loss to why this is happening.  Thanks for any help.

The first issue to look out for, is to make sure that
you mix and match the correct versions of OS, matlab,
and MEX function.

A 64-bit OS can run either 32-bit or 64-bit matlab.
A 32-bit matlab can *only* run 32-bit MEX, whereas
a 64-bit may or may not be able to run 32-bit MEX
in addition to 64-bit MEX.

Your compiler may or may not be able to generate 64-bit
object code.

Make the slightest mistake in this mess, and you will
likely see the kind of errors you observe.

So, to have half a chance to port your code:

1) Find out what version, 32-bit or 64-bit, of matlab
you run (presumably 64-bit).

2) Find out whether your compiler is able to generate
64-bit object code (it might not be).

3) Find out what compiler flags and switches need to be
set to generate 64-bit code (not done automatically)

4) Set the appropriate flags and switches, and re-compile
the whole mess from scratch.

Rune
From: Taylor on 18 Jun 2010 07:04
"Steve Amphlett" <Firstname.Lastname(a)Where-I-Work.com> wrote in message <hvfg1d$rrv$1(a)fred.mathworks.com>...
> "Taylor " <twsouthwick(a)byu.net> wrote in message <hvfc0g$grj$1(a)fred.mathworks.com>...
> > I have written a mex file that compiles fine within 32-bit windows xp and works without any problems. However, when I tried compiling it under a 64-bit linux environment, I get the following error after compilation (which occurs just fine):
> >
> > Mex file entry point is missing. Please check the (case-sensitive)
> > spelling of mexFunction (for C MEX-files), or the (case-insensitive)
> > spelling of MEXFUNCTION (for FORTRAN MEX-files).
> > ??? Invalid MEX-file
> >
> > I've searched for why this happens, but I can't seem to find anything useful. My project is in C++ and I'm compiling it with g++. In Windows, if I compile it with visual studio, it works just fine. I have the entry point spelled correctly with the correct arguments. I'm at a loss to why this is happening. Thanks for any help.
>
> First place to start would be a symbol dump from the mexa64 file:
>
> % nm whatever.mexa64
>
> to find out what symbols it actually does have inside it.

The correct function is:

void mexFunction(int nlhs, mxArray *plhs[], int nrhs,
const mxArray *prhs[]);

but I hadn't been using the const for the last argument. It would compile under visual studio, but not under g++. It compiles and works just fine now.
 | 
Pages: 1
Prev: Matrix to Column vector
Next: Use of uigetfile() together with uiwait()