On Mon, 28 Jun 2010, Paul Kerry wrote:
> One of our users has noticed a problem when using object detection under
> gaia (32 bit linux hawaiki) along with fits format files: they are
> object detecting and then saving the resulting detected objects to a new
> file to view - they assure me that this worked in the previous nanahope
> release. I've tested this and the same happens for me, although this
> doesn't seem to affect sdf files which are opened and then object
> detected.
>
> Open a fits format image (either a saved one or one downloaded from a
> catalogue) in gaia
>
> image-analysis -> object detection
>
> under the checkimage tab:
> checkimage type -> detected objects
> check image name -> test.fit
> (no other parameters have been changed from the initial defaults)
>
> then click "detect objects"
>
> when test.fit is opened in gaia, the background pixel values are zero
> and the detected objects are in the range of ^-41
Hi Paul,
thanks for the detailed report. I can confirm this problem under 32bit,
which isn't present in older builds or the 64bit one.
I've had a look for the problem, which is in SExtractor, but haven't
managed to find it, or make it go away, so I'm afraid this will have to
remain unfixed for now.
I suspect this is a compiler issue, probably a bug in the code that has
just been revealed, so maybe updating SExtractor itself to the most recent
version would help.
If this is a urgent case I'd recommend just switching back to the old
version of the "sex" binary if you can.
Cheers,
Peter.
|