Re: [ferret_users] Ferret crash with shade

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Bruno,
I have found the cause of this bug and it will be in the next Ferret 
release.

I have a comment about this question.  I know that a SHADE plot is a 
common way to have a look at data. I do that myself all the time. Of 
course the image, which has several hundred pixels in each direction, 
cannot capture the full resolution of the variable on a grid with many 
thousands of points. So it's good to be aware that what we're seeing is 
a decimated version of the full variable. (Ferret plots all the data, 
but some pixels will "overwrite" each other as the plot is made.) You 
will get just as much detail and be able to control how the data is 
decimated, by using strides,

yes? shade votemper[L=1:82000:50]

or regridding it to a coarser grid.

Ansley

On 9/16/2013 10:14 AM, Ansley Manke wrote:
> Hi Bruno,
> Thank you for the report.  This happens with older versions of Ferret 
> as well. (I was wondering if it might be some detail in 
> double-precision Ferret, or something that occurred in a particular 
> version of the NetCDF libraries, but it seems not.)  It does not seem 
> to relate to coordinate storage; the same thing happens if I force the 
> coordinates to be stored as if they were regularly-spaced.  For 
> regularly-spaced grids, we store only the first coordinate, coordinate 
> count, and delta-coordinate.
>
>    yes? use/regulart mooring.nc
>
> Your last example, where FILL looks so odd, is I think a good clue. A 
> fill plot with up to 50000 time points is fine, over 50000 fills in 
> part of the field with green. This seems to me that there is an 
> unintended fixed value somewhere, perhaps in the graphics code, and 
> which has different symptoms when the SHADE command is used.  I will 
> see if I can track this down.
>
> Ansley
>
> On 9/13/2013 5:01 AM, Bruno Levier wrote:
> > Hi !
> >
> > Using shade causes Ferret (v6.82) to crash with a specific dataset 
> > (in attachment).
> >
> > yes? use mooring.nc
> > yes? sha votemper
> > **ERROR Ferret crash; signal = 11
> > Segmentation fault
> >
> > yes? use mooring.nc
> > yes? sha /L=1:82000 votemper ! OK, but ...
> > yes? show data
> > Bus error
> >
> > It doesn't crash with fill, but the result is not good :
> >
> > yes? use mooring.nc
> > yes? fill /l=1:50000 votemper ! Seems good
> > yes? fill votemper ! Not good
> >
> > Perhaps is it due to the limit of storage for coordinates ?
> >
> > Thanks in advance for any help.
> >
> > Bruno Levier