Supported Software: MOPEX Bugs

Known bugs in MOPEX

1. PROBLEM
   In the current APEX 18.1.5, when the switch "Use_Photerr_for_SNR = 1 or 2" is on, there are sources that mistakenly get a "bad" SNR value of -10.0.
   
   SOLUTION
   To be fixed. For now don't set "Use_Photerr_for_SNR" (or set = 0). For SNR, it will default to an estimate based on the point-source profile (see readme/README_new_snr_option_081108.pdf). For a better estimate of point-source signal-to-noise, use an aperture uncertainty, where the aperture covers the main core of the point-source.



2. PROBLEM
   Using the 18.1.5 apex.pl, when the switch "use_data_unc_for_fitted_SNR = 1" is on, and there are tiles with no data.  APEX sourcestimate will exit with a message "number of noise images less than number of tiles".
   
   SOLUTION
   The Perl script has been fixed (Version 4.4) and is available here: apex.pl. Replace the one in your MOPEX download sub-directory /bin with the new one.



3. PROBLEM
   I'm trying to start the MOPEX GUI on my Linux box, but I keep getting the following error on startup:
   
   

   	prompt> ./mopex
   	Error occurred during initialization of VM
   	Could not reserve enough space for object heap
   	Could not create the Java virtual machine.

   
   
   SOLUTION
   The MOPEX GUI startup settings are too aggressive for your machine. You need to edit the following line in the startup script (<installation_dir>/mopex):
   
   Replace the following line:
   set MEM = (-Xms48m -Xmx2048m)
   
   with this:
   set MEM = (-Xms48m -Xmx1024m)
   
   then run the script to start MOPEX. MOPEX will prompt you to auto-update. Select "Quit and Install" to update to the latest version. After MOPEX is updated you must repeat the above steps to change the startup parameters of the newly updated startup script. You should then be able to start the MOPEX GUI without a problem. This issue will be addressed in future updates, and the fix should not have to be repeated.



4. PROBLEM
   When running a simple pipeline with Drizzle interpolation and no outlier rejection, I get an error when I try to run the following work flow:
   Initial Settings -> FIF -> Mosaic Interpolate -> Mosaic CoAdder -> Mosaic Combiner
   The error reads "You have set up Mosaic Interpolate with 'Interpret Method = Drizzle'. This setting depends on the output of Mosaic R Mask. You must add Mosaic R Mask into your flow."
   
   If I add the required modules, it still fails because Mosaic R Mask does not create any files if no outlier rejection was run.
   
   SOLUTION
   Create a bunch of empty masks and specify these as the RMasks in "Intial Setup". The pipeline then runs fine with no error messages.



5. PROBLEM
   When specifying the Fiducial Image Frame file FIF.tbl, the file must be in the output directory. The name of the file can be changed, but must have the extension .tbl. If no .tbl file exists in the directory then MOPEX will not even open the directory for viewing.
   
   SOLUTION
   For now, ensure that the FIF.tbl file is located in the output directory before attempting to select it. It can be named anything, so long as it has the extension .tbl.



6. PROBLEM
   When running APEX on my Linux machine, it fails in the sourcestimate module with an error about glibc. Note: the following fix has also worked for some users with a range of glibc problems encountered when running versions of Linux other than RedHat. As with all Spitzer software, however, we give no guarantees about running APEX on any non-RedHat Linux platform.
   
   SOLUTION
   We have confirmed that the current Mopex sourcestimate module can fail on Linux RedHat WS 4 with a glibc error. The binaries were built and tested on the RHL equivalent to RH WS 3. The version of glibc shipped with RH WS 4 provides some additional checking and it is objecting to some aspect of the memory allocation in the code, which we have not yet identified.
   
   Right now, we believe it is not a serious error, but that glibc is being too stringent. As a temporary work-around, there is a way to shut off the stringent checking. Before running, in a tcsh:
   
   prompt> setenv MALLOC_CHECK_ 0 (0=be silent, 1=show error msgs, 2=abort silently, 3=abort, show msgs.)
   
   Note the underscore at the end. There is a discussion of this under the heading "glibc" in: http://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/release-notes/ws-x86/. Be sure to exit the shell after your MOPEX session is done if you want other code to run under the full glibc.



7. PROBLEM
   There is a problem with the FINERES option in the Mosaic Interpolation module, as it doesn't handle masks right.
   
   SOLUTION
   No solution yet. FINERES is an option in the default interpolation scheme. It is used if present and set to a non-0 value in a namelist, or has a non-0 value set in the Mosaic Interpolate Settings box in the GUI. We recommend not using FINERES if you also need to use masks.



8. PROBLEM
   I am using the MOPEX GUI on a mac. Sometimes, when I load a namelist which includes the APEX pipeline, the "Select" Settings Dialog will not launch.
   
   SOLUTION
   Check the paths to the data in the namelist. In particular, check 'INPUT_FILE_NAME', 'OUTPUT_DIR', and 'PRF_file_name'. In a terminal, go to the data directory and type 'pwd'. Make sure that the paths in the namelist match the output of 'pwd'.
   
   Sometimes external disks can be accessed in various ways, and the GUI cannot understand all of them. For instance, to access an input file on an external disk by clicking on the "Add" button in the GUI, you might first have to click on "Macintosh HD". If you write out the namelist after doing this (using File-->Save As...), the path will include "Macintosh HD". If you then reload this namelist, you will find that your "Select" Settings Dialog will no longer launch. Editing the path will fix this problem.

Fixed in MOPEX v18.1.5 (08/21/08)

1. PROBLEM
   The recommended PMask and DCE Status Mask Fatal Bit Patterns for IRAC have been updated, but these changes are not yet reflected in the GUI IRAC reduction templates.
   
   SOLUTION
   Make sure you are using the correct Fatal Bit Patterns by checking the "pmask" and "imask" links on the IRAC pages at: http://ssc.spitzer.caltech.edu/irac/products/.



2. PROBLEM
   Exported APEX namelists mistakenly uses the full path to specify PRF file(s). Command-line MOPEX (e.g. apex.pl) requires that the PRF be in the cal/ directory.
   
   SOLUTION
   Put the PRF file(s) in the cal/ directory. Edit the exported namelist file to remove the pathname.



3. PROBLEM
   In Mosaic R Mask there is an error checker that wants the Bottom Threshold less than the Top Threshold. This makes no sense since these values are on opposite sides of the mean and you may want to be less conservative on the positive "top" side than the negative "bottom" side.
   
   SOLUTION
   Ignore the error message.



4. PROBLEM
   Displaying a large stack of images is very slow. In the Initial Setup module, you have the option to display only a selected range of images, but that is not yet implemented for the other outputs. For now, be aware that if you have >>100 images, the display will be very slow.
   
   SOLUTION
   None at present - the ability to display a selected range of images will be added at a later date.

Fixed in MOPEX v16.3

1. PROBLEM
   The list of input files cannot be changed after running (part or all of) a flow. Even going back to Initial Setup and changing the definition of the input list doesn't work; the GUI will remember the old one.
   
   SOLUTION
   To use a different set of input files, start a new flow enitrely. To retain the current settings, export the current pipeline to a namelist file and then read it back in.



2. PROBLEM
   When running APEX Multiframe in the GUI, it fails in the module sourcestimate with the following error message:
   
   ERROR: SOURCESTIMATE: number of masks is not equal to the number of input images in /Users/csb/Software/Mopex_GUI/mopex/cal/mips24_pmask.fits
   
   Surely I should only have 1 pmask input file? What's going on?
   
   SOLUTION
   This is a bug in the GUI, only affecting APEX Multiframe. It will be fixed in the next update of MOPEX, due to be released in early November. In the meantime, you can use the command-line version of MOPEX to run APEX Multiframe. You can set up the namelist from within the GUI and export it for use on the command line.



3. PROBLEM
   Exported APEX namelists mistakenly write out the keyword PRFMAP_FILE_NAME = PRF_Map.tbl even if no such file exists. If there is no such file, the command-line apex.nl will not run.
   
   SOLUTION
   Edit the exported namelist to comment out that line if there is no PRF_Map.tbl.



4. PROBLEM
   There are some slight differences in the output of the GUI version of APEX and the command-line version. These typically only affect a small fraction of the measured objects (< 1%). These differences occur in both Fit Radius and Source Estimate and can lead to moderate differences in output parameters (at the < 10% level for the few affected objects). In some cases, the slight difference in parameter values can lead to an object be chosen by Select in the GUI and not on the command line.
   
   SOLUTION
   Fit Radius is wrongly using the output of MedFilter Detect, rather than the output of MedFilter Extract. This will be fixed in the next major update, but in most cases the differences will disappear if you use the same parameters for the MedFilter Detect and MedFilter Extract modules.



5. PROBLEM
   The Create RMask Mosaic module hangs when the input stack has more than ~25 images.
   
   SOLUTION
   The mosaic is usually created, so the flow can be aborted at the end.
   
   Alternatively, the mosaic pipeline can be exported to a namelist and run through the command-line MOPEX. To limit the command-line to just Create RMask Mosaic, turn off all of the run_* switches, except for create_rmask_mosaic. For example:
   
   

   	have_uncertainties = 0
   	compute_uncertainties_internally = 0
   	run_fiducial_image_frame = 0
   	run_mosaic_interp = 0
   	run_mosaic_proj = 0
   	run_mosaic_covg = 0
   	run_mosaic_outlier = 0
   	run_mosaic_rmask = 0
   	run_mosaic_reinterp = 0
   	run_mosaic_coadder = 0
   	run_mosaic_combiner = 0
   	create_rmask_mosaic = 0




6. PROBLEM
   The Aperture module allows the value of N Apertures to be any integer, but the dialog only lets you input 3 aperture sizes. If N Apertures is greater than the number of input sizes, the module will crash.
   
   SOLUTION
   Set the value of N Apertures to 1, 2 or 3.

Fixed in MOPEX v16.2.8

1. PROBLEM
   A bug in Select causes a loss of precision in some output columns. Importantly, the precision of RA and Dec is reduced to 0.36 arcseconds.
   
   SOLUTION
   The values in *_extract_raw.tbl are unaffected. Either use the raw table only, or use the source ID numbers in the Select-ed table to find the original values in the raw table. This bug will be fixed in the near future, so be sure to check for updates.



2. PROBLEM
   Exported APEX namelists do not put Select conditions in quotes, so the command line does not read them.
   
   SOLUTION
   Edit the exported namelist to put quotation marks around the Select conditions.



3. PROBLEM
   Both Pixel Ratio and Pixel Size are written to the namelist file when a Mosaic flow is exported. This is a problem, because the absolute pixel size will override the pixel ratio if both are present - even if the pixel ratio was selected in the GUI's Initial Setup, the exported namelist will use the absolute size.
   
   SOLUTION
   Edit the exported mosaic namelist to have only Pixel Ratio or Pixel Size before running the command-line mosaicker.



4. PROBLEM
   After running Initial Setup, the review of settings lists the pixel size as both absolute and by ratio.
   
   SOLUTION
   There is nothing to be done at this time. The GUI will use the choice that is input to General Options before Initial Setup is run. It just won't display which choice was made in the review of settings.



5. PROBLEM
   In the command-line MOPEX, the Mosaic Outlier module allows a threshold of zero. This is useful for having flexibility in setting the threshold later in the RMask module instead. In the GUI, the minimum outlier threshold is 1. In general, this should not cause any loss of flexibility, because RMask can still choose any threshold >=1, and outliers are rarely defined at <1 sigma. However, when reading a command-line namelist, a threshold of zero will generate an error.
   
   SOLUTION
   After reading in the namelist, set the threshold settings in the GUI to 1.