Task and Tool Documentation, getting this into a standard
Update of data repository
Build my tasks on OSX
Sandra: Run MPI tests before release, Akeem- these are being tested.
There was no meeting last week.
Technical Meeting tomorrow 15:00 UT
Justo: Enable stack trace code in pre-release.
CASA 4.3.1 / EVLA pipeline commissioning: intent handling and ASDM version issues
CASA 4.4. / interferometry pipeline testing: no new problems, plotms, i/o changed still to be tested
CASA 4.4. / HPC testing: bottleneck in EVLA pipeline, time not spent in CASA tasks
development: support for line lists in the imaging tasks, web log
Building / Installing and parallelization
Testing the ALMA pipeline with a larger data set containing 3 EBs. It fails in lustre due to an openMPI bug. I am at the moment testing the patch put together by Julian.
Many meetings and discussions.
worked on resolving EVLA pipeline intent handling issues with Brian Kent
worked on resolving problems with multi-vis task QA handlers with D. Muders and S. Williams
completed QA metrics for hif_exportdata task
explored options for generating scriptable image lists for ARCs
Tested Justo's fix for writing flags in plotms, resolved CAS-7393
Ran plotms tests to check for any other issues
Known issues for release 4.4: virtual model data (CAS-7390) and improving averaging VIVB2 performance (CAS-7412)
Got the green light from Jeff to commit plotms in the trunk. Let the games begin!
Dynamic linker error on OSX 10.10. Emailed Darrell.
Fixed a bug in single dish gridder for multi ms griding
Fixed msuvbin when gridding with wprojection on to get flux correct
ran FC20 alma-m100 test at the request of Dirk on an eso server (have to be a bit more proactive about OS we support..took more that 1/2 a day just to download tar ball and setup stuff to get the run going.).
Justo González Villalba
Investigated together with Julian system call problems in MPI env., it turned out to be a problem with OpenMPI signal handling.
Fixed problem in back-propagation of flags in MSTransformIterator when input channels are dropped
Large number of meeting
Susan Loveland: Almost entire week spent on Vewer/Profiler CAS-5689 (No spectral fitting support for ellipse and polygonal regions). Very little CARTA work.
(CAS-7405) rg.fromtexfile uses more than 64GB for 1000 circles. Fixed by using a broad and shallow data structure rather than a narrow and deep one. Also substituted smart for raw pointers in related code and compute composite region on demand rather than as lines are added from region files, which means storage of intermediate data structures is no longer necessary.
(CAS-7408) immoments crashes with underscores in mask names. Kicked back to reporter as this is likely the result of not escaping special characters in LEL properly (as per CASA Note 223).
(CAS-7186) specsmooth doc typo. Fixed.
(CAS-5689) spectral fitting results seem off. Investigated and punted to Susan as this seems to be the result of the viewer using the rectangle circumscribing the ellipse as the region, rather than the ellipse itself.
(CAS-7204) ia.regrid no longer works on certain images in the new release. Bug was the result of the rectangular image approximation being invalid in this case, since the images are dozens of degrees on a side. I removed the direction coordinate overlap check if images are > 1deg on a side.
Removed MSMetaData compile warning flagged by Ger.
(CAS-6971) imfit logger messages should indicate noise scheme being assumed. Done.
(CAS-7414) msmd.almaspws(chavg=True) returns True for FULL_RESOLUTION spws that have been channel-averaged offline. Fixed.
(CAS-7370) beamArea in viewer statistics needs more general calculation. Worked with Nathan on testing of this fix.
Reported tAipsIO test failure to Darrell.
(CAS-4890) test_ia_subimage[test_stretch] leaves open image. Fixed by replacing raw with smart pointer fields in casacore ExtendImage. The raw pointers were not being deleted when exceptions were thrown during construction of ExtendImage objects which ultimately caused the subimage mask image to be left open, as well as associated memory leaks.
(CAS-7394) Investigate issues with ImageInterface ->cloneII(). Resolved as Won't Fix since what I thought was an issue with cloneII() was really a symptom related to CAS-4890.
(CAS-7227) better reporting of tabular spectral axes. Punted to Dirk since this part of the imhead output is generated by CoordinateSystem::list() in casacore.
CAS-5223,4469: enchanced initweights functionality to permit a bit more flexibility in initializing weight and weight_spectrum. Existing sigma or weight column can each be used as the weight info to initialize with.
More discussions with James re gaincal performance; began testing with his (big) dataset.
Re-started work on freq-dep statwt (CAS-6062,6064; v4.5); initweights work provides a useful framework for providing this.
Attended ngVLA workshop.
(CAS-7225) mosaic clean in alma-m100-analysis-regression fails on Fedora Core 20: working with Kumar to understand reasons for failure; got account for Kumar; for him the test passes; for me it fails; suspect setup problem
(CAS-7416) Perform ALMA HPC tests with realistic conditions and setup: working on preparing a large performance test with ALMA data cailbration; identified datasets; now adapting script to work with stable and running in batch
alma-m100-regression: set up automated run on EL6 at ESO with test builds; email will be sent to casa-staff in case of failure
(CAS-7421) Make the distances parameter of im.fixvis available in task fixvis (and make it work): looking into this at the request of Todd and Bryan
Started implementing correct non-natural imaging weight calcs for parallel imaging runs (cas-7005).
Looked at code to see how to also address cas-7210 and do correct sensitivity calcs.
Got mpi4py going on laptop for imager. Discussions with Justo/Sandra about controlling log messages from mpi4py so that imager's logs are readable...
backported openmpi signal handler fix, solves random system() failures (CAS-7419)
Continued working on cube imaging partitioning for tclean - added optional 'csys' and 'imshape' to the image parameters and changed buildCoordinateSystem() to use that coordsys instead of building one from other image parameters when the parameter is defined