Monday Morning Meeting 11 May 2017

  • DIAL-IN NUMBERS & PASSCODES:
  • IP: 192.33.117.12##8110
  • Phone: (434) 817-6524

Minutes in blue

Attendance

  • Socorro: Juergen, George, Martin, Kumar, Urvashi, Tak
  • CV: Anand, Akeem, Ville, Bob, Andy, Jen
  • Garching: Sandra, Federico, Enrique
  • SCO:

News / Meetings / Visitors

  • Will the CASA meeting be moved back to Monday? Anand will pass question along to Morgan.

Build, Release

  • Branching today
    • Jeff says we are ready to branch.
    • Potential blockers:
      • CAS-10086 -- Memory issue and performance improvements; Kumar says the fix is in.
      • CAS-8809 -- chanchunks; an offshoot of autoboxing; Kumar will look at the dataset today. Not a branch blocker.
      • CAS-8530 -- model column; Kumar says fix is in.
      • CAS-8588 -- Kumar says fix is in.
    • When we branch:
      • svn commits will be turned off; everything will go through git.
      • We will have a 5.0 branch and the prerelease (master)
      • In Charlottesville and Socorro, we can use the -r option to start a particular version of CASA
    • Don't expect pull requests being merged until early next week
  • Bamboo upgrade went okay. Lost DB connection about a week ago; after a reboot it has been working fine.

Verification Testing

  • NGC3256 regression test has been placed in to quarantine
    • Should be the last failing test in the Bamboo accepted list
    • Iterating with the NAASC DAs, the problem seems to be in gaincal, applycal, or clean
    • The issue has been lingering so long, finding the history has been a challenge
    • It has never worked in the new Bamboo system
  • CASA guide mediawiki upgrade
    • Josh is upgrading mediawiki today (concluded at 9:46 ET) - was it a success? TBD
    • casaguides were in read-only for about 45 minutes
    • Will casaguides be updated to CASA 4.7, or will it be postponed to 5.0 release?
    • Who will be posting the updated extract script in the guides?
  • importvla is now in the framework
  • buildmytasks test

Validation Testing

  • status update on CAS-10086, CAS-8809, CAS-8530, and CAS-8588? as far as I know these are holding up branching. (see Build, Release, above)
  • Trying to knock down the biggest 5.0 tests: autoboxing and VI/VB2 checks
  • Andrew tested polarization; sent comments to George; George already checked in fixes.

Architecture

  • Warnings Summary
    • Note that although the alma warnings count went down as expected with the shift to devtoolset-4, there are now 12 warnings for display that someone may want to look in to.
    • Script may not be using the right flags; Ville thinks this is not urgent; maybe resolved with movement to bamboo.

HPC

  • Will parallelization be included in the 5.0 release? Not clear.

Development

  • Interesting performance slowdown in running gencal for VLA quantizer. Cache in ms metadata not large enough somehow.

Pipeline

AOB

Developer Reports

Monday Meeting
  • Sanjay Bhatnagar
  • Sandra Castro
  • Lindsey Davis
  • Pam Ford
  • Enrique Garcia
  • Bob Garwood
    • work on tests of asdm code generation move in the ALMA code tree
    • sick for 2.5 days
  • Kumar Golap
    • VI2 fixes; mosaicing stuff for VLASS
  • Jeff Kern
  • David Mehringer
  • George Moellenbrock
    • Fixed a VI2 logic problem in blcal
    • Fixed a few tasks that were still using global tools in whole or in part (which now doesn't work): imporevla, blcal, uvmodelfit)
    • CAS-9098: Further study of (rare) gencal slowdown for EVLA RQ---turns out many MSMetaData methods are filling its internal cache with large whole-MS column objects, and for the particular MS reported in the ticket 12.499M rows, leaving too little space in a 50MB (~=4*nrows!) for the time/field map, and so it gets re-calculated within a busy loop in the RQ parsing code. The answer was not to increase the cache (which would just cause datasets of a different size to be frustrated in a similar way (not to mention horrible waste of memory). Instead the cache in the gencal context is now made to be slightly smaller than the size of a scalar Int column. Dave will look at refining the caching conventions in MSMetaData.
    • CAS-10118: Fixed a problem with VI2 in ALMA Poln wherein the VB2::correlationTypes() values were inadvertently changed (in VisBufferImpl2.cc) when using smodel in solving tasks.
  • Dirk Petry
  • Martin Pokorny
    • Completed initial design for asynchronous VI2, started write-up
    • [VLA: realfast]
  • Federico M Pouzols
    • Looking into subtle bug in flagdata with rflag and combinations of correlations CAS-10036.
    • SynthesisImagerVi2 - VI/VB2 setFrequencySelections violating asserts in some (corner?) cases: CAS-8809.
    • Runnig tests of pipeline and tclean for CAS-9519 and CAS-10086.
  • Urvashi Rao
    • Autoboxing
  • Darrell Schiebel
  • Ville Suoranta
    • Investigated and fixed Carta/Codesign issues
    • Fixed CAS-10112 buildmytasks path issue
    • Started to work on getting the warnings to Bamboo
    • Investigated and fixed various build issues
  • Takahiro Tsutsumi
    • A minor fix on automask processing order (CAS-9208)
    • Testing on the fluxscale modification for robust flux bootstrapping (CAS-9186)

Friday NAOJ Meeting
  • Kanako Sugimoto
  • Wataru Kawasaki
  • Masaya Kuniyoshi
  • Takeshi Nakazato
  • Renaud Miel
Topic revision: r13 - 2017-05-17, JaredCrossley
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback