Re: segmentation fault and netcdf/mex problems - R14S3/Red Hat <long>
- From: "Nikki" <nprive@xxxxxxxxx>
- Date: 9 Dec 2005 10:17:44 -0800
> The MEXEXT command gives the extension MEX files need to have to be
> recognized on your system. For 32-bit Linux, MEXEXT should return mexglx;
> for 64-bit Linux, if I'm not mistaken, it should return mexa64. Therefore,
> on machine B, your MEX file will need to be named netcdf.mexa64 to be
> recognized. You shouldn't just rename the MEX file, though -- I'd recompile
> it or see if the location where you obtained the library before has a 64-bit
> version of this file.
Ah, yes, Computer B would like mexa64 files instead. I will look for a
64-bit version and have this recompiled.
> > This is very reproducable, but only on Computer B (yes, I should send
> > this to support, all 20 files involved). I do not get a segmentation
> > fault if I print a simpler figure with the same command, only when I
> > try to print my complicated figure (which has an imported jpg
> > background overlain with several thousand line segments). For example,
> > if I plot only the background image, then use the same print -djpeg
> > command, Matlab prints with no problems.
> Since you have a figure with which you can consistently reproduce the
> failure, please send it in to our support staff as a FIG file -- they will
> let you know if they need the code to recreate the figure. Also, as a
> potential workaround, try changing the Renderer property of your figure
> before printing. I believe, from the stack trace, that it's set to
> 'zbuffer' -- try setting it to 'painters' or 'opengl' and see if the
> segmentation fault still occurs.
I will send in the FIG file to support. I tried setting the renderer
to both painters and opengl (by adding '-painters' or '-opengl' into
the print command list); neither of these worked.
> > Are these problems possibly related? Has the netcdf library become
> > corrupted in the transfer from Computer A to B, or could there be a
> > deeper mex-related problem? Is there any other way I can output a jpg
> > with a set size and resolution aside from the print command?
> Probably not; probably not, I think you're trying to use a MEX file for
> 32-bit Linux on 64-bit Linux; and look at the IMWRITE function,
Does the IMWRITE command support assigned resolution? I need to output
a jpg which is 4096x2048 in size, and have been using the combination
set(gcf,'PaperUnits','inches','PaperPosition',[0 0 16 8]);
to get the right output size. The options for the jpg type in
'imwrite' do not include resolution.
I have noticed that Matlab sometimes gives a segmentation fault when I
exit or quit on the new Computer B (by entering "exit" at the command
line) - the program looks like it is quitting normally, but the
terminal window gives a segmentation fault. This is erratic and I am
not sure yet if there is a combination of events which makes it
Thank you for your help,
- Prev by Date: Re: A slick way to generate this matrix?
- Next by Date: Re: Passing output variable names into a function
- Previous by thread: Re: segmentation fault and netcdf/mex problems - R14S3/Red Hat <long>
- Next by thread: loop to access array elements randomly