doing a lot of ad hoc testing of tclean through the imaging PL benchmark testing
10.11 testing update (?)
CASA testing prioritization meeting
Had a false bit of joy when the warnings scoreboard went all green, but on further examination I had a stale version of casacore and it killed the actual build. It's still pretty good, though, with only the code/alma having significant numbers of warnings.
Google test info: Put up a page with an example of using Google Test. I'll expand that this week.
Unit test status: We still have a large number of old unit tests which haven't been integrated with the CASA build system. It's likely that the tests still work. Just take a look at them to see if they are building and passing and maybe relevant; if so, then change them over to to the new cmake macro ( instructions here). The display and singledish projects have all there tests integrated!!! (No credit given to projects with 0 of 0 tests integrated as that would be mathematically unsound, at the minimum).
Effective C++11 book: Scott Meyers has a new book out called Effective Modern C++ which has some pointers about using the features in C++11 and C++14. The NRAO has a site account to Safari books, so NRAO employees can read it online (from a work computer) for free.
Crash Reporter: Waiting to make sure we don't have any possible exploits
Julian created some patches for the MPI sub-communicator prototype. After several discussions, Martin has successfully used these patches on his tests. We should have a telecon soon to discuss the outcome of these tests and what to do next. On this subject, Julian has asked Kumar and Darrell for comments on how to add a PyObject wrapper class so we can pass through a PyObject to the python exposed tool functions. Could Kumar and Darrell please have a look at this problem and get back to Julian?
Commented on Kana's testing of the baseline separation axis in partition/importasdm. Added the baseline option also to mstransform's interface. Ticket is resolved (CAS-7890).
Started to test SD pipeline with the baseline axis. My preliminary tests show that the whole processing is ~ 13% slower when MMS is partitioned per baseline compared to 'auto' partition (balance of spw and scans). Will discuss it with SD team at some point.
patch required to deal with caltable / FIELD table / ephemeris table and fluxscale gainfield='nearest' issues
patch required to deal with ALMA TP data observing mode changes: missing plots, mosaic phase center calculation, Tsys spw map
HPC out of memory problems in hif_findcont task, seem to be in Python and / or CASA tool calls not tclean
EVLA validation and acceptance testing
EVLA low SNR bandpass heuristics
Discussion of image / primary beam image coordinate systems
Discussion on how to tell if ALMA science target is a mosaic
New Tuesday EVLA pipeline coordination meetings
Users may be looking at the ca tool again
Issue in nearest and ephemeris information
Definition of MS -V2 does not require that any data specified in the Subtable must exist in the MS.
Started to implement a temporary change to the MMS structure (CAS-7896). In short, only the SOURCE sub-table will be duplicated in all Sub-MSs. All other sub-tables will be linked to the one in the first Sub-MS. For that, I'm making changes to the tasks that write to the HISTORY table, so that only the MPI Client writes to it when running in parallel. The final changes to the MMS structure will come after setjy is refactored.
Investigated ALMAC3R4 pipeline gainfield='nearest' solve and flux scaling issues and provided George with test data
Investigated and provided patches for CASA 4.5.2 / pipeline C3R4 problems, plotms, gainfield=nearest
Selected data set for new ALMA pipeline acceptance test, and worked with Brian on details
Completed ALMA low SNR gaincal heuristics
Investigated and closed old tickets (cannot reproduce or duplicate) - CAS-2384, CAS-2530, CAS-3667, CAS-6042
CAS-8197 Not a Bug (ALMA feed/scan/spw selection issue)
CAS-8165 Not A Bug (plotms BSOD) - looks like user clicked "Close"
Followed up Jurgen's science testing of generalized pre-averaging in flagdata.
Reviewed MPI communicator Python->C++ transfer protoype.
Fiddling with the design of a subtable framework to provide access to the subtables (probably through the VI) without letting aspects of the table-system leak out. This is useful in that it hides the details of the table system from the user and provides a more convenient way of accessing the information. It also makes it easier to synthesize subtable information on the fly for use in testing or chained transforming VIs.
CAS-8319: Segfault when changing between optical/radio moments with no data loaded
CAS-8382: Profiler & Moments should ask the user if ther want optical or radio version if it has not been specified previously.
CAS-8249: Viewer crashes when opening another new panel from an empty panel.
Meetings (2.5 days)
Bug fixes & investigation of shared library build problem for Alex.
Redesigning code to take into account groups of layers.
CAS-8313 investigations. Eph info is not propagating to caltables, so gainfield='nearest' is mis-matching Eph objects. Additionally, this is triggering a deeper problem, apparently related to reuse of solutions over multiple MS fields. Investigation continues....
cas-6937 : clarity in what kinds of restarts are supported in tclean
Several minor things since the last meeting ( mask regions partially outside the image, turned on widefield unittests, added more tests, went over a number of old tickets to evaluate and close them or send for testing, fixed non creation of .pb for runs with chunking,... )
Spent most of the time on looking into tclean(+makepb) error (CAS-8355)