weird issues with gpcal, uvplt, mirflag

[sarahr]

Hi,

I've been trying to reduce some of my CABB data in miriad.

I was initially reducing the data on my laptop (Mac OSX 10.9, binary installation of miriad - latest version), but found that a couple of the tasks were exiting before completing (ie. the task starts running fine, but returns to the prompt part way though without giving any error message).

In trying to determine the cause of the problem I discovered some very weird behaviour. The tasks that were causing problems on my laptop ran without fault when I tried them on Linux. I also found one task that ran fine on my laptop, but didn't run properly on linux (same problem - task exits before completing).

A summary of the affected tasks (and which operating systems they do/don't work on) is given below. This isn't a complete list, just what I had access to to try.
The tasks with problems don't fail every time though. With mirflag I perform an initial flagging step, followed by bandpass calibration, then more flagging. On linux mirflag works fine before bandpass calibration, but fails afterwards (but works both before and after on Mac). On Mac gpcal and uvplt still work on some sources but not on others (but work on all sources on Linux).

I normally run my data reduction using a python script (with mirpy) but have also tried doing the reduction manually from the miriad shell, with the same result. I also tried downloading the original data files again, in case the data were corrupted somehow, but this made no difference either.

I have previously reduced this same data, without problem, on my old laptop (OSX 10.6) with an earlier version of miriad (not sure what version). There may have been very minor differences in the reduction method, but nothing significant. I don't have access to a machine running 10.6 to test with the latest version of miriad though.

The data can be found here:
http://www.physics.usyd.edu.au/~sarah/miriad_errors/

For each task above I've included a datafile to run the task on, as well as a text file giving the exact inputs I've used, and the resulting output (on the operating system that caused the problem). Where possible I've also included an example of a file where the same task runs fine (on the same operating system).

I have no idea what's causing this so any help would be much appreciated!

Cheers,
Sarah.

[ste616]

Hi Sarah,

Thanks for the report. I'll take a look at the data tomorrow, but this message is to let you know that we no longer recommend that people use mirflag. The pgflag task provides a more capable automatic flagging routine and should be used instead of mirflag.

[Mark.Wieringa]

Hi Sarah,

I had a quick look at the gpcal issue and I think it may be related to the bandpass solution - according to the history you used interval=1,1,1 in mfcal, which makes a new bandpass every minute - with 117 bandpass solutions you may be hitting a memory limit (the program crashed when about to apply solution 116). If you really intended to set the bandpass solution interval, you could try something like 1,1,60 and see if that works better.

Cheers,

Mark

[sarahr]

Hi Jamie and Mark,

Thanks for you quick replies.

I tried re-running the reduction with the interval for mfcal unset and that seems to fix the issues I was having with gpcal, and also uvplt

The issue with mirflag is still there (on linux) post-pandpass (but since it works fine on my laptop, this isn't a major issue to me). Thanks for the advice on pgflag - I'll look into using that from now on.

Cheers,
Sarah.