CASA User Testing Meeting Minutes - 2015-09-15, 09:00 Socorro / 11:00 Charlottesville
NOTES ONLY: Meeting canceled.
- Socorro: AOC-317; IP: 18.104.22.168
- Charlottesville: Room 331; IP: 22.214.171.124
Minutes (Main Points Only)
- 4.4.0 has been released for all supported platforms.
- 4.5 code freeze was scheduled for August 17th 2015. The 4.5 pre-release branch was created at this point. Test/stable/monthly packages are also still being produced.
- The same general-purpose CASA launcher script has now been deployed at CV, SOC and GB. Details are on the appropriate drop-down menu of the public CASA pages at http://casa.nrao.edu
- For anyone using test machines zuul01, 03 and 04 are running RHEL6, zuul02 is running RHEL5. REMEMBER: Use /usr/bin/casa on these machines to make sure that you are running the new test/stable packages (there are RPM installations on these).
- 4.3.1 ALMA and VLA Pipeline version released (RHEL and OS X links posted). Pipeline major release number expected to catch up with the rest of CASA at 4.5.
- Planned 4.5 release timeline:
- Aug 3: Feature Freeze (all new features are now in the codebase and ready for test) <- PASSED
- Aug 14: Code Freeze (Bug fixes etc are expected to be in the trunk at this point, we should have all regressions passing as well) <- PASSED
- Aug 17: Release Branch Created <- PASSED
- Feature tests complete: obviously detected issues will need to be fixed and retested, but everything should have been at least had one look
- Aug 17-31: <- PASSED
- Identified defects are being fixed, documentation generated
- All commits require authorization
- First round of GUI testing
- Aug 31: First release candidate available for random user testing <- PASSED
- Sept 9: Release Target <- PASSED
Please conduct all remaining 4.5 user testing with 4.5 pre-release packages whenever possible.
Number of CASA Jira tickets with "Ready To Test"/"Under Test" status (including EA, EU, JAO, Software-only tickets, but excluding Pipeline) prior to review: 60.
Number of CASA Jira tickets with "Ready To Test"/"Under Test" status (including EA, EU, JAO, Software-only tickets) following review: 43.
- 4.5 Blockers
- New ticket assignments.
- The "nag list", "input required" items
- Major test coverage areas for 4.5 testing
- tclean testing to be handled as a short-term project all of its own. Strategy was agreed upon for ALMA between Jeff & Crystal.
- VLA tclean testing report from Claire:
- gaincal gets different results when I use tclean with savemodel='virtual" vs. savemodel='modelcolumn', with the virtual model apparently worse. If plot both models in plotms as amp vs. uvdist they look the same, but there must be something different about them as far as gaincal is concerned.
- Help documentation for savemodel seems to be a placeholder, needs to be completed.
- Unclear what the parameter "overwrite" does based on inline and other help. One would imagine that overwrite=T would overwrite the existing images, but in fact is uses them and starts from the existing model, which is not at all clear from the documentation.
- Dedicated parent tickets with systematic subtasks specifically to track tclean testing efforts have been set up: CAS-7664 and CAS-7669. The main questions to be addressed will be: (1) Does the user interface work as expected? and (2) Does it actually do the right thing?
- An ALMA Pipeline testing package intended purely for internal NRAO testing use only has been put together that makes use of tclean. It is expected that the Pipeline people in Europe will be comparing manual and the new basic automatic imaging. The PA Pipeline people will be performing the "deep dive" testing (e.g. terms:2 at B3, etc.).
- For the 4.5 CASA release, it is anticipated that one or two CASA Guides will be produced that essentially outline "This is how you do it with the old clean, and here is the tclean equivalent."
- It will be necessary to ensure that the clean regression tests are shown to work with tclean (modulo any appropriate tweaking).
- Crystal et al. have access on lustre to a CASA 4.5 test/stable package containing tclean that she was to begin testing from June 29th. The overall aim is to at least ensure that it works well enough for ~5 or 6 of the most common-use cases.
- Crystal previously reported that there were actually three different implementations of the Clark clean algorithm in CASA, and due to contrasting results when attempting to clean mosaics and single fields, two of these have now been linked in for tclean (old Clark -> "clark2" in tclean (suitable for single field); Clark -> "clark" in tclean (suitable for mosaics)). Using the inappropriate one appears to generate deep negative holes in images, and it is suspected that the two implementations were over-optimized for specific cases. Steve had also noted that tclean seems to exhibit problems with restarts (e.g. first run with a mask, then second without).
- Parallelization is also expected to feature more prominently in CASA 4.5. It would be good if this could be relied upon to help speed up flagdata, although our most immediate concern is that it is not responsible for additional breakage (particularly if any such breakage makes it more difficult to test/debug any tclean issues).
- For a limited group: HPC work (e.g. number of open files, etc.). The Pipeline group are also involved in testing this. This probably does not need a lot of additional dedicated science user testing yet, beyond routinely trying to use split2, day-to-day MSTransform-type work, etc. Amanda had been looking into testing routine split2, cvel2 and other MS_Transform-related issues.
- Ephemeris handling
- Weight column (including weight column -> initialized weight spectrum). See CAS-4469. Testers to work with George. Steve was to look into this.
- Other matters previously arising
- From Claire: plotms: plotting the virtual model column results in a message ?: MTFTNew :: fromRecord" being printed on the terminal but not in the log. It should probably be in the log only (if at all, as it's not obvious to the user what it means).
- It had previously been decided that a wiki page would be created for the NA Imaging Team to collect together a list of common use cases to be tested. Socorro staff were then to review the list, and add any VLA-specific entries for which they also see a need. Juergen was to provide a copy of his old plotms testing automation targets to use as a starting point.
- BDF flagging has been shown to be problematic. Fixes are being put in by Michel, but it has been an ongoing issue. See PRTSIR-6800.
- Anand was to ask Josh to upgrade the remaining DA OS X 10.8 Mac to 10.9 for user testing purposes.
- Mark R. really needs to get going on the tools documentation again soon. The documentation on the imager (in particular) is just plain wrong in places.
PART 1: Items to be Covered During Meetings (as time permits)
4.5 (Non-Pipeline) Blockers
- CAS-7664 tclean science user testing: Continuum (MFS) <- tclean parent ticket. Urvashi suggests a review of the 4.5 blocker status for this ticket, as a lot of work will be needed... Discussion still ongoing. Pretty close, though.
- CAS-7831: plotms no longer correctly plots the virtual model. Under Test. Tester: Catarina. Virtual model stuff generally flagged as a possible cause for concern. VI/VB interactions? Testing currently ongoing.
- CAS-7856: viewer image display is strange in 4.5.0-REL. Scheduled. New 4.5 blocker item. Darrell is investigating.
- CAS-6629: Data/Image partitioning scheme for Cube Imaging. Currently Scheduled, but it sounds like work has been completed on this and committed to pre-releases and trunk. May be resolved once CAS-7862 has been successfully tested.
"Ready To Test" Items Pending Tester Assignment
- CAS-7744 (Jul): 4.5 item. Baseline-based types unflag data on apply. Under Test. Former blocker, now demoted. Tester: TBD (was Crystal); Fix confirmed to work for blcal. Still needs to be tested for the chan-independent cross-hand phase case. Need a tester with a suitable data set for this.
- CAS-7820: plotbandpass(overlay='antenna,time', showatm=True, field = '') fails to show the atmospheric curve if final Tsys was not done on one of the specified fields. 4.6 item. UT. Tester: DA (ticket current still with Anand).
- CAS-4239: plotms(iteraxis='baseline') should display the baseline length in the default title of the plot. 4.5 item. tester: DA (Rachel?). Ticket with Anand.
- CAS-7778: msmd should have a method to retrieve the effective bandwidth or the effective resolution of the channels rather than only their formal bandwidth. UT. Tester: TBD. (provisionally Juergen)
- CAS-7637: msmd should have methods to retrieve REST_FREQUENCY and TRANSITION from the SOURCE table. UT. Tester: TBD. (provisionally Juergen)
Propose to Close Items
Items Overdue for Status Updates (i.e. the "Nag List")
- CAS-7862 (Sep 2015): tclean segfaults with parallel=True. Tester: Stewart W. Once this has been tested, can probably resolve blocker item CAS-6629 as well.
- CAS-7726 (Sep 2015): plotms shows data that it should not. 4.5 item. Tester: Juergen
- CAS-7699 (Jul 2015): plotms transform=True produces an incorrect x-axis scale when avgchannel is also set (even to 1). RTT Tester: Had been Mark L. (and Amanda). High priority testing item! Tester: Amanda
- CAS-7607: 4.4 plotms GUI testing Mac 10.10. 4.5 item. RTT. Tester: Rachel H.; Juergen to also do some testing. Ongoing?
- CAS-6130 (Jul): 4.5 item. Parametrized tables in cal library. Under Test. Blocker. Tester: Steve. Following recent discussion, provisionally pushed back to be a 4.6 item, as Steve doesn't think it constitutes a blocker for a 4.5 release.
- CAS-3080 (Jul 2015): flagmanager mode='save' -- default behavior should not destroy existing flagversion. 4.5 item. UT Tester: Crystal. Looks good. Just awaiting a Cookbook entry before resolving.
- CAS-7315 (Jul 2015): Migrate phase shifting code to TVI. 4.5 item. Subtask of CAS-4864. MSTransform issue. Mostly now as expected, but some oddities still seen, one fix committed, commit from fixvis dev needed. Tester: Amanda.
- CAS-7109 (Mar 2015): Linear full mosaics. Available in stable now. Under Test. 4.5 item (former 4.4 item. Testing started, Input requested from developer. Fix put in by Kumar for problem found. Tester: Steve. Ticket pinged repeatedly.
- CAS-4983 (Mar 2015): frequencies in gain tables derived using combine='spw' are incorrect. 4.5 item (formerly 4.4). Available in stable now. Under Test. Tester: was Claire, now Crystal. Some duplication with CAS-7097 (assigned to Claire). Claire reports: "Have tested gaincal with solving for individual spws, looks OK, labels for delays now match complex gain, at least in plotcal plots (used to be edge frequency instead of center frequency). Still have to test with averaged spws." Crystal had felt that Claire's testing so far was already almost sufficient. While the remaining testing should be done before the release, it is probably less urgent than other issues right now. Unlikely to get final testing done for 4.5 release now. Call it good enough, or push out remaining testing work to 4.6? * CAS-1390 (Aug 2014): ImageRegrid required for Complex images. Under Test in 4.5 item (was 4.4). Tester: Preshanth. Ticket pinged repeatedly. * CAS-7314 (Mar 2015): Add capability to average across fields. Officially Under Test already, but needs the VI/VB2 version of PlotMS to test, so this would be an advance assignment. Intended science use case for this: Different observations of the same field source might carry different field IDs. May also have mosaicking applications. 4.5 item (was 4.4). Tester: Steve. Ticket pinged repeatedly.
- CAS-3953 (May 2015): Subtraction / Fixing of outlier fields in clean. Subtask of CAS-6622. Ready to test in tclean via interactive viewer. VLA item. Minor issue. Tester: Steve. Ticket pinged repeatedly. * CAS-5524 (Jul): 4.5 item. import uvfits should do make consistent when getting visibilities in STOKES form. Under Test. Tester: Kumar. Ticket pinged repeatedly.
- CAS-6215 (Jul): 4.5 item. data averaging in flagdata. Under Test. Tester: was Emmanuel, passed to Steve. Ticket pinged repeatedly.
- CAS-7718 (Jul): 4.5 item. setjy to use 'Perley-Butler 2013' as default standard. Under Test. Tester: was Juergen, passed (via Todd) to Bryan B. Ticket pinged.
Other Items Currently "Under Test"
- CAS-4663: specfit components/prompt/inf. 4.6 item. Under Test. Tester: Juergen
- CAS-4735: imsmooth needs additional option for single dish use. Subtask of CAS-5844. Under Test. Tester: Amanda
- CAS-7562 (Sep): 4.5(?) item. Write model to MS in parallel tClean. Ready To Test. Passed to Jeff for reassignment (probably to one of the parallelization folk?).
- CAS-3366 (Sep): 4.6 item. listfile parameter for imstat. Under Test. Tester: Juergen
Single Dish (to be skipped during the meeting)
- CAS-4469 (Jun+ 2013): Split with channel averaging is creating spurious spectral features and weight is wrong when there has been channel flagging. Subtask of CAS-5245. 4.5 item. Critical item. was "Under Test". Extensive comments provided by George and Justo. Currently "Input Required" from Crystal. Ticket pinged repeatedly.
- CAS-5583: Frequency in exportuvfits output. Currently with Brian K., who is liaising with the user in question. No recent feedback from the requesting user, though.
- CAS-5806 Mathematical prediction of post-gridding single dish beam. Formerly a 4.3+ Blocker, now "Critical". "Input Required?* Developer has done some modelling, but... ? This is a "blocker of a different kind..." High priority (blocking further development). Crystal and Kumar had been in discussion, but this ticket has been quiet since (effectively) December. 4.5+ item now. Currently Input Required from Daniel.
- CAS-6124: imfit subtasks:
- CAS-4538: migrate imfit input estimates from file to input parameters. Input Required (was "On Hold"). Subtask of CAS-6124. Ticket passed to Juergen, pending a CSSC discussion of what needs to be done and a scheduling of the work. Pushed back to 4.6. Ticket pinged. Juergen: More thought about imfit still needed by Science.
- CAS-4748 (Dec 2012): Polarization selection does not select SIGMA. Input Required. Passed to Jeff for prioritization. Sanjay thinks it's a low-priority item. Version 4.6+.
- CAS-5886 (Dec 2013): bizarre clean boxes for large images. Awaiting input from one developer to another. Version 4.6+.
- CAS-5162 (Sep 2013): Improvements in flagging documentation and examples. Input Required from Science (via Jeff). This ticket has been left untouched since May 2013... Version 4.6+.
- CAS-6794 clean is very slow handling this simple, small ACA mosaic. Input Required from Juergen. Kumar commented on this. Ticket pinged
- CAS-7599 imfit deconvolved peak flux unit. Input Required from Juergen (and other CSSC members).
- CAS-6688 Turn on new vi/vb within imager and test. Was a 4.4 item, but now Unscheduled. Currently with Jim. Some scope modification suggested for this ticket. Confirmation received that this is not actually a 4.5 release blocker.
General questions had previously been raised about callib expectations for 4.5. Implementation in PlotMS? (See Drew's ticket, above). Scratchless operation? Mark R. has asked George about this on CAS-7096. More progress than we suspected has been made on the latter, but is expected to be regarded as "experimental" for the purposes of a 4.5 release.
- CAS-7407 CalibratingVi2 Improvements. Currently Scheduled". 4.6 item (not a 4.5 blocker). Ticket with George M.
- CAS-7287 Deploy Calibration-aware MSTransformManager in solve context. Subtask of CAS-7096. 4.6 item. Currently Scheduled". Ticket with George M. Ticket pinged.
- Viewer registration tickets:
- CAS-6560 contour image does not unzoom when master coordinate image is re-registered. Subtask of CAS-6604. Viewer issue. Currently Unscheduled. Not listed as a 4.5 blocker.
- CAS-6720 2 panels don't register. Viewer issue. Currently Unscheduled. Not listed as a 4.5 blocker (but continues to embarrass...).
- For additional (non-user testing) interest, see also:
- CAS-2609 Factor out code of Conv. func. cache management. Nominally a 4.5 item. (Fix in, some testing still to be done by Sanjay, but appears to work)
- CAS-7861 CASA dies on VLA ephemeris datasets. Scheduled. No fix version specified, but being worked on.
- CAS-7835 importasdm() does not handle VLA observations of ephemeris objects properly. Scheduled. No fix version specified, but being worked on.
- CAS-7760 code and asap atnf libraries clash. Scheduled. No fix version specified, but was being discussed in August.
- CAS-7755 tclean produces strong negative residuals. Scheduled. No fix version specified, but was being discussed in June.
PART 2: Items Listed for Informational Purposes (note that these will not normally be specifically discussed during the meeting unless someone raises them)
- tclean testing (for info only: to be skipped during the meeting: will be handled by the Imaging Group)
- CAS-5048 (May 2015): Clarify when peak residual calcs use/ignore masks, and use consistent thresholding rules. Older ticket, now ready to be tested using new tClean in 4.5. Tester: TBD (was Emmanuel, back in the 4.1 cycle). Ticket with Urvashi
- CAS-3729 (May 2015): clean to write out only those images that are necessary for a given run. Subtask of CAS-4640. Tester: TBD. Ticket with Urvashi
- CAS-3290 (May 2015): Logging : clean logging for cube cleaning needs improvement. Subtask of CAS-6615. Tester: TBD. Ticket with Urvashi
- CAS-4133 (May 2015): clean inline docs for masks and new region format... Subtask of CAS-6617. Tester: TBD (Brian Kent, who has history with this ticket? Or DA?). Ticket with Urvashi
- CAS-1623 (May 2015): CLEAN should have option to continue from previous CLEAN state. Subtask of CAS-4640. Tester: TBD. Ticket with Urvashi
- CAS-7292 (July 2015): clean spw parameter cannot stride >1 channel. 4.5 item. Under Test. Tester: Crystal (or other member of imaging team).
Unable to Test Items
- CAS-4758 (Sep): 4.4 item. fluxscale fails if one of the reference fields is flagged. Previously "Under Test" (r31121+) in current (new-style) stable. Tester: Claire. No testing progress on this since October. Claire had been trying to find a suitable dataset for testing. Steve thought that this should actually be fairly easy to test and that Claire and/or he may have been able to test this one. Ticket with Brian Kent. Ticket pinged. Currently On Hold, pending test data availability.
- CAS-5379 - Subtle inconsistencies with other programs in Galactic Coordinates (Adam and Harvey). In a nutshell, shifts of ~0.2" are seen when taking data into other packages (e.g. DS9, IDL which both seem to agree with each other). It is unclear yet if this is an incorrect implementation in one or more packages, or possibly just a precision error. This is a subtle issue, and was pushed back. Ticket stalled. Needs to be looked at by someone with lots of time. To be left as-is for now. This needs significant work, including new science papers on how to actually define the Galactic coordinate system? Needs casacore discussion; this can be part of that. Passed to Jeff for now. Pushed back to some indefinite future release. Unscheduled.
- CAS-4636 (Oct 2012): Elevation dependent opacity application w/ Tsys. Still awaiting input from Science. Had been "On Hold" since August, awaiting data. Previously switched to "Input Required" and passed to Crystal for comment. Bumped back to 4.4+. Work on this is Unscheduled until further notice (more than a year, now).
- CAS-3532 (Sep): Incorporation of Alma Primary Beam in Mosaic. Currently Unscheduled. Tester was Scott S., but ticket passed to Crystal for input. Some test imaging done. Ticket with developer. Won't be done for 4.5 now.
Upcoming Ready to (Re)Test Items
Not Yet Assigned To Testers
See below "assigned" SD list.
Already Assigned To Testers
- CAS-4592: plotms on 10.7 issues. Previously half fixed due to Qt updates. Shown to till be an issue under 10.10. Currently Unscheduled. Tester: was Nathan, to be Juergen. Exported plots all look OK, even if the plots in the GUI appear corrupted. Juergen to discuss prioritization with Jeff? Fix in time for 4.5 now seems unlikely...
- CAS-5893 (May): Now a 4.5 item. Add the ability to write a spectral profile to an ASCII file in the ia tool. Was "Under Test" in 4.3. Now Unscheduled. Tester: Assigned to Juergen for testing and a possible CSSC discussion. Might also need a task adding for this, but this will be 4.6 work now. Recent testing done, passed back to developer for 4.6+ work.
- CAS-5127 (May): Document ia toolkit methods. Status Scheduled. Mark should start working on this again... Most of Juergen's prioritized list of 26 items had been done. Dates to be added his future documentation updates. Ongoing. ia.regrid docs will require more work. Still needs code commits. Mark to restart work on this under Linux soon.