iraf.net

I am using IRAF V2.16 and SAOImage DS9 6.2. I have problems with ELLIPSE: the task works, but I cannot see the ellipses overlaid on the greyscale image displayed by DS9. In fact, I cannot use ELLIPSE interactively. On the contrary, when I run ISOIMAP, the ellipses appear on the display. Could you help me, please? The operating system of my laptop is Fedora 16.Best regards,
Elena[color=red:706e3c93c3][b:706e3c93c3]Posted on behalf of a user by moderator.[/b:706e3c93c3][/color:706e3c93c3]

Yes, I have exactly the same problem. In fact, I think ELLIPSE is
indeed drawing something, but all at x=0, because I can see a red
vertical line at the edge of the display (i.e. at x=0). (ISOIMAP also
works correctly, as Elena notes, but then I can't work out how to tell
ISOIMAP only to show the values in a table, and not interpolate.)I'm also using IRAF v2.16, STSDAS 3.14, ds9 v6.2. (I tried
ximtool v2.0 as well, but that didn't help.) I wonder if this could have
anything to do with running some of the code in a 64-bit
environment. I only suggest this because I have the identical problem
on _two_ machines, both running the same versions of IRAF and OS: kubuntu 11.04, 64 bit (x86_64)However, on my laptop -- kubuntu 11.10, 32 bit (i686) -- ELLIPSE works fine, and draws the contours as we know and love. Unless a solution to this problem has been found, perhaps we should
contact STScI, since they're the ones that construct STSDAS/ELLIPSE?CheersDave

Thanks for the post, it gives me a few ideas: I don't think it's specifically a mixed-architecture problem, the CL will happily use a package's 32-bit binary when using a core 64-bit architecture. However, the CL in this case is still the 64-bit version and is responsible for providing the graphics kernel to render the plot. There've been issues in the past where graphics metacode was found to be incompatible due to changes in the size of an INT, and this may be a case where the 32-bit-generated metacode isn't being handled properly by the 64-bit CL.To isolate this, you could try running a 32-bit CL explicitly, i.e. invoke the $iraf/bin.linux/ecl.e from the commandline to start a CL (or else temporarily swap the normal $iraf/bin.linux64/ecl.e with the 32-bit version and start normally). If you use a 32-bit CL and ELLIPSE works then either the CL has a bug or the ELLIPSE task is doing something not seen elsewhere to set the window boundaries. It would also help to know the date of the CL binary you are using, or else do a "make latest" from the $iraf directory to be sure you're running the latest update.If you get a plot, could you do a ":.write <file>" and ":.gflush" to save the metacode to "<file>" and post that to the anonftp at ftp://iraf.noao.edu/pub so I can try to reproduce the problem here? OTOH, if you see the same problem then it may be an issue for STScI to take up.

Hello again. Thanks for the quick reply... I tried the tests you
suggested. Here's what happened.1. I did a "make latest" in $iraf. No change in the behaviour of
ELLIPSE.2. I downloaded iraf.lnux.x86.tar.gz (i.e. the 32 bit
version). Although you already know this, I thought I should make a
note of how the bin files are configured before I messed with them. My
default build produces: /irar/iraf/bin -> /iraf/iraf/bin.generic -- empty
/iraf/iraf/unix/bin.linux -- empty
/iraf/iraf/unix/bin.linux64 -- not empty, contains ecl.e
2a. I went with your second method to replace the 64-bit ecl.e with
the 32-bit ecl.e. (Trying to run the 32-bit version outside of IRAF
caused all sorts of errors when trying to start IRAF.) So I did the
following (...and I've left out all the "sudo" commands btw):% cd /iraf/iraf/bin.linux64
% cp ecl.e ecl.e.SAVED # hang on to the original 64-bit version
% cp ~/dump/32bitIRAF/bin.linux/ecl.e . I went back to my IRAF directory to start IRAF. And viola! ELLIPSE now works!I wasn't sure what exactly you wanted a plot of (".write" doesn't work
in ELLIPSE?) but I'm happy to provide whatever you need.
So I guess the one remaining question now is, what to do next? Is
there any difference between running ecl as 32- or 64-bit? Should I
just leave the 32-bit executable where it is and get on with it? Would
I see any difference anywhere?CheersDave

[quote:b940ade3eb]
I wasn't sure what exactly you wanted a plot of (".write" doesn't work
in ELLIPSE?) but I'm happy to provide whatever you need.[/quote:b940ade3eb]The ":.write" is one of the standard graphics cursor commands, my guess is that either you forgot the ':' or you were in the 'image cursor' mode at the time. Is the failure in drawing a graphics plot or overplotting the image display? Can you upload some sample data/parameters to the anonftp at ftp://iraf.noao.edu/pub so I can try to reproduce the problem?

We had a problem running ellipse using the 64-bit version of 2.15.1a; we got a segmentation violation (and "illegal instruction") whenever we used it in interactive mode so we were interacting with the image display. Isoexamine had the identical problem. I upgraded to V2.16 64-bit and reinstalled stsdas and the problem went away.