Hi Peter, William,
thank you for the detailed report, and for the sample file in the Debian
bug report. The effect also happens with the (debianized) version 8.3b1
which is in Debian Experimental. I don't think this directly connected
to WCS: The output we see here is
%1.8G
which looks much like outputting a format string instead itself instead
of the formatted string (see attached image). When one changes the
setting for "Precision --> Coordinates --> Linear" from 8 to 5, the
output changes to "%1.5G", which shows that it is really the format
string, built from the precision setting. However, I still was unable to
find the place in the code where the format is used to print the output.
William, can I ask you to point me to that location in the source? I
promise that I will do the rest of debugging myself :-)
Best
Ole
On 23.05.21 18:45, Peter Teuben wrote:
Hi Bill
thanks for some useful info on what goes on behind the scenes in ds9
I now use 8.3b1 actually, and as long as I have your binaries it
never bit me. It was when I was using a virgin laptop with just
ubuntu/debian packages. And I could not easily upgrade my ubuntu to use
8.2 or 8.3, i have to do some fiddling with some apt setting, and given
that my laptop was effectively bricked after yesterdays disaster, I'm in
no mood to experiment on other machines.
But Ole has the file and debian debug report, and he could see if 8.3b
is also effected by this, i.e. if it's a WCS issue. Or maybe I need to
learn to write a better WCS :-(
peter
PS: my laptop bricking could also be a hardware error, I'm not blaming
the firmware update and/or the debian bug reporter that seems to have
de-installed some dell specific things. It's an SSD, but the health
reporter only claimed 3% of the spares were taken.
On 5/23/21 12:10 PM, William Joye wrote:
All
The current official release is v8.2. It is based on Tcl/Tk 8.6.10.
I’ve also released v8.3b1, which is based on Tcl/Tk 8.6.11.
I can’t really say why the change in WCS behavior, other than we are
talking about binaries over 2+ years old. The current version uses AST
8.7.1 for all WCS support. I no longer try to ‘fix’ screwed up WCS
definitions. You get what the FITS file creator defined, screwed up or
not.
Also, lots of 2D data out there with 3D WCS defs, along with 3D data
cubes with 2D WCS defs. DS9 will blindly display what AST calculates,
as ‘David Berry’ knows best :-)
Also, DS9 needs OPENSSL 1.0.x, while the default for most ports now is
1.1.x. I don’t know how Debian/Ubuntu is handling this. In my
binaries, I compile 1.0.x and statically link.
Peter, you should upgrade to at least 8.2 or 8.3b1, as it has so many
bug fixes and new features, including Dark Mode and Prism.
And Ole, thanks for maintaining the DS9 ports for Debian and Ubuntu.
Bill
On May 22, 2021, at 10:02 AM, Peter Teuben <[email protected]> wrote:
hi Bill
i usually grab the ds9 binaries from your site (now via
google), but ubuntu also offers a package "saods9". Sadly it has a
different approach, which is killing some of my WCS application.
Maybe Ole knows the provenance of this.
It runs /usr/share/saods9/library/ds9.tcl directly via wish. Yours is
a true executable, at least as far as I can see.
For most RA-DEC-FREQ stuff I use, this is not an issue, but for some
more non-standard slicing of cubes I wind up making my own CTYPE's:
CTYPE1 = 'RADIUS ' /
CTYPE2 = 'VELO-LSR'
CTYPE3 = 'Z ' /
and this works fine on yours , I can read off the WCS values. but for
the ds9 from ubuntu/debian (still at v8.1) I get to see the value
%1.8G
in the top left panel where the WCS is "named".
And one more item about WCS I wondered about.
Incidentally, the labeling of the axis, where I see "FK5" for our
cubes, I get to see "RADIUS-Z-SPEC..." the last word is cut off
because there is not enough space. But it's SPECTRUM. And I wondered
why Z was mentioned before SPECTRUM.
If I make the cube a 2D image, so take the "Z" axis away, the label
reads "RADIUS-SPECTRUM"
But if I change VELO-LSR to PHI, it correctly reads R-PHI in my
case. So i guess VELO-LSR is the SPECTRUM axis, but then why was Z
listed before SPECTRUM earlier when I had a 3D cube with the Z axis
only 1 in length....
best,
- peter
