Started with implementation of denoising lib, based on GSL. Goal is to provide thread-safe lock-free generalized linear squares, weighted linear squares and re-weighted (robust) linear squares routines. Also high-pass, low-pass and median filters will be included. I will use the new UVContSubTVI implementation as a test bench. The details on the various re-weighting strategies to be discussed in a PDR to be held at the end of October.
Helped debug a problem that showed up on the RHEL 6 build of CARTA. Turned out to be a static initialization problem that did not show up on Ubuntu builds where development builds are performed.
Fix of timing segfault on RedHat6 initialization (Helpers:Ville found problem & provided RedHat6 debug environment; Jim diagnosed problem)
(CAS-6896) immath should give control over file used for header generation: IR. Discussion of the many potential issues related to the imagemd parameter and how much hand holding (= added development cost) the code should do in order to alert users they are doing stupid things.
(CAS-8929) imstat(axes=[0,1,2], listit=False) still lists the per-channel statistics: Resolved. Not a bug. verbose=False controls this behavior (yes I too find it annoying that there are two parameters in imstat that control the amount of logging; not my doing).
(CAS-8928) when using beam ia.convolve2d() throws error that minor/major/pa are specified even if they are not: IR. The default values are "specified" (I don't necessarily understand why, because the values don't make much sense, but not my doing). This should be able to be resolved by explicitly specifying empty strings for the relevant parameters.
(CAS-8918) hif_importdata is stuck (python CPU bound for 16 hours) at the "Analysing" step of the 5th of 6 ACA7m Cycle 3 science ASDMs: WIP. Iterating with Julian on PR. It was a caching issue.
(CAS-8783) add parallel support to imstat/ia.statistics(). UT. Responded to test team questions.
(CAS-8837) implement multithreading in ClassicalStatistics::_createDataArrays(): UT. Responded to test team questions.
Read new process documentation.
(CAS-8901) msmd needs a feature to retrieve spw IDs from spw names: RTT. Implemented msmd.spwsfornames().
CAS-8993: checked this fix manually into release-4_7, since pull #485 not yet merged to casacore:master.
CAS-8904: communications w/ Takeshi re test_tsdcal failure (test data update req.). Also discussions re precision w/ Kana: Basics: Interpolate1D.h nominally supports mixed types, but uses ArrayMath operators which do not (i.e., casacore not supporting itself). A proposal is forthcoming to remedy this.
CAS-8913: Turns out migration to (~indirect) use of MSMetaData to set scan numbers for ALMA Tsys is sabotaged by inaccurate timestamping of Tsys values by the telescope; reverted to ad hoc way which builds in ALMA lore to get it "right".
CAS-8923: A user reports "numerical instability" in D-term solve for ALMA data for cross-hand phase near zero. I've found and fixed (release-4_7) a problem in the cross-hand phase solution which is likely the cause.
Structure101 repository now has a reasonable dependency graph snapshots, and I am working on daily snapshots and incremental updates.
4.7 branching and related build configuration
Investigated Carta build issues on RHEL6
Investigated/fixed Carta packaging issues on Mac
Worked on separating branch and master build/packing workflows on Bamboo
Configuring builds on the new Mac VMs
Investigated CAS-8584. Asked Kumar to change the interpolation used in SetJyGridFT to linear to make the setjy results in a better agreement between multiple spw ran separately and merged together. Made some minor changes to setjy code as well. All checked in to both the release branch and trunk.
Added a tolerance to divergence check in tclean so that it stops as "diverged" only if the peak increases more than 200% between major cycles ( CAS-8584 )
Friday NAOJ Meeting
An issue in ATM library found ( ICT-7997): could we fix it in 4.7?
I could finally log in to Plone. I believe all developers (except for Masaya, who hasn't tried) can log in now.
assigned starter development targets of 5.0 to developers who are not working on new developments.
reported a couple of issues found in PL and other tests: plotting SD caltables by plotms ( CAS-8892) , an issue in nearest frequency interpolation of LSB spws in applycal ( CAS-8893 fix done. waiting for fix being pulled), intent selection issue in calmode='otfraster' (no ticket)
investigating tsdcal failure on 4.7 prerelease: due to the fix in CTTimeInterp1, need to update reference data