CASA Testing log Alpha release. S.T.Myers 26 October 2006 Search through this report for "NOTE" for some encountered problems and comments. running on sandrock: /home/sandrock2/smyers/Testing2/Oct06/20061026 this file: 2006-10-26-myers.txt _______________________________________________________________________________ Fire up casapy: > source /home/ballista/casa/daily/casainit.csh /home/sandrock2/smyers/Testing2/Oct06/20061026 CASA Version 2.0 Build 125 Thats strange, how come this build is lower number than the 143 in the /home/casa version? (Joe says he saw that too and has Wes on it.) > casapy casapy: error while loading shared libraries: libCCM_Python.so.0.5.5: cannot open shared object file: No such file or directory I cannot run the daily from sandrock (nor imager_b). Other machines give sometimes different errors. Note this is one of the reasons Joe wants to distribute via rpms. However, since the /home/casa version works, then it should be possible to serve a daily version, say from /home/casa/daily/ Joe is working on fixing this problem. For now, log into basho and run there. From basho: > casapy Try mosaic again (Joe says he fixed this). > inp 'mosaic' > imagename='testmosaic1.dirty' > mfalg='mfclark' > imsize=[2500,2500] > nchan=1 > spwid=[0,1] > vis='/home/sandrock2/smyers/Testing2/Oct06/20061025/am693_20010701_cband.ms' > fieldid=[2,3,15,16,17,29,30] > mosaic exceptions.TypeError Traceback (most recent call last) /home/sandrock2/smyers/Testing2/Oct06/20061026/ /home/ballista/casa/daily/linux_gnu/python/2.4/mosaic.py in mosaic(vis, imagename, mode, mfalg, niter, gain, threshold, mask, nchan, start, width, step, imsize, cell, stokes, fieldid, spwid, weighting, rmode, robust, scaletype, constpb, minpb) 133 im.setvp(dovp=True) 134 im.setdata(mode=mode,nchan=nchan,start=start,step=step,fieldid=fieldid) --> 135 im.setimage(nx=imsize[0],ny=imsize[1],cellx=casac.Quantity(cell[0],'arcsec'),celly=casac.Quantity(cell[1],'arcsec'),mode=mode,nchan=nchan/width,step=width,fieldid=fieldid) 136 im.weight(type=weighting) 137 im.setoptions(ftmachine='mosaic',padding=1.0) TypeError: an integer is required Oops. The problem is that fieldid is being used for the list of fields in setdata and for the phasecenter in setimage. The phasecenter will have to be a separate parameter since you need to allow for fieldid=-1 etc. Ideally, would allow for a direction also (like it could be between field centers). Joe has set it up for now to be the first element of fieldid (as in the clean task). He is working on putting phasecenterid in as another parameter. Eventually that will just be phasecenter which can take either an integer id or a direction. Note that it didn't get far enough to write a mosaic.last file :( Have to exit and reenter casapy to pick up Joe's fix. > fieldid=[16,2,3,15,17,29,30] > mode='mfs' > saveinputs 'mosaic' > mosaic Many iterations with Joe omitted as it was clear that the daily was in fact using older versions of stuff. This worked. Looking at the mosaic in qtviewer shows that my mosaic apparently suboptimal spacing, though I thought the FIRST guys set it up correctly. It may also be due to the mosaicing kernel in the mf implementation which effectively multiplies by an extra primary beam factor (Kumar confirms this). Need to divide by the normalization image (Kumar is adding the output of this to the task). It looks like there is indeed a source in field 2. Now wait for Joe to rebuild the daily to fix the update problems. ---------- 2006-10-27 ---------- > source /home/ballista/casa/daily/casainit.csh /home/sandrock2/smyers/Testing2/Oct06/20061026 CASA Version 2.0 Build 125 Even though the version is the same (and wrong) Joe says it updated. Indeed this version of mosaic has the new phasecenterid. Lets try a bigger mosaic 2 3 15 16 17 29 30 42 43 44 56 57 69 70 71 83 84 96 97 98 > execfile 'mosaic.saved' > imagename='testmosaic2.dirty' > imsize=[2500,2500] > fieldid=[2,3,15,16,17,29,30,42,43,44,56,57,69,70,71,83,84,96,97,98] > phasecenterid=43 > saveinputs 'mosaic','mosaic2.saved' > mosaic Note that this takes a long time even on basho. And it crashed (segmentation fault). > casapy Seems to hang - basho problem? Yes, seems hung (or connection problem). Maybe related to Joe trying to fix tarzan... Wait for it to come back. > execfile 'mosaic2.saved' > mosaic mosaic -- OptionError: '['']' is not an option for 'stokes'; must be one of: ['I', 'IV', 'IQU', 'IQUV']. Apparently the saveinputs incorrectly saves the stokes. > stokes='I' > mosaic Segmentation fault (core dumped) OK, that failure was real. Lets try the original mosaic again > execfile 'mosaic.saved' > fieldid=[2,3,15,16,17,29,30] > phasecenterid=16 > saveinputs 'mosaic' Note stokes is saved incorrectly stokes = "['']" > mosaic This worked. Note mosaic.last has incorrect stokes also. Must be that the 20 field case had memory issues (though no error in casapy.log). Looks like it may have crashed making the residual image. Try fewer fields. testmosaic1 was 7 fields. Do 13 15 16 17 29 30 42 43 44 56 57 69 70 71 > execfile 'mosaic2.saved' > imagename='testmosaic3.dirty' > stokes='I' > fieldid=[15,16,17,29,30,42,43,44,56,57,69,70,71] > phasecenterid=43 > saveinputs 'mosaic','mosaic3.saved' > mosaic This worked, but mosaic image looks centered at the wrong place, like it actually is using the first entry, not phasecenterid. (Joe confirms he hadn't completed the move to phasecenterid. My bad.) > fieldid=[43,15,16,17,29,30,42,44,56,57,69,70,71] > saveinputs 'mosaic','mosaic3.saved' > mosaic Yes, this is correctly centered. There are nice source between fields 43/70 and 42/69. Note you now get a .fluxscale image (courtesy Kumar). This clearly shows the strong weighting pattern, which drops to 0.4 from 0.9 max in the interstices between mosaic centers. Clearly not optimal! NOTE: as noted by others, having to draw a rectangle and double click outside it to decrease zoom is not a good way to unzoom. Needs to be an offzoom button (to restore to fully unzoomed view). Currently, unregistering ALL the images and then re-registering restores the unzoomed image. > execfile 'mosaic2.saved' > stokes='I' > fieldid=[43,2,3,15,16,17,29,30,42,44,56,57,69,70,71,83,84,96,97,98] > saveinputs 'mosaic','mosaic2.saved' > mosaic OK, this worked (wall time was about 4min). Clearly the crash was due to centering the mosaic at the corner. NOTE: bug in imager (probably) when the phasecenter is sub-optimally chosen such that it crashes. Looks good. Note that testmosaic2.dirty.image is 64MB. Would like to try some cleaning. Unfortunately, mosaic (and clean) only take masks at this time. Could use the ia tool to make some masks, but we really do need the "makemask" task (see UIWG report). NOTE: Even better, we should be able to make a text boxfile with lists of BLC,TRC (four integers) or CENTER,RADIUS (three integers) for rectangular or circular clean boxes respectively (also see UIWG report). This would greatly aid automapping scripting and also would be an easy way to do semi-interactive clean (look at viewer, note source positions in pixels, add box to file with position and radius or blc,trc). Nominally could get box coords from qtviewer, but right now it doesn't display the pixel coords of cursor. Of course, I can just go into the Wrench menu and set: Position Tracking->World or Pixel Coordinates=Pixel and it now shows this, but it is sufficiently useful that I would think that it could always be shown. NOTE: qtviewer should also display the pixel coordinates of cursor in tracking window at bottom. Note bright sources are at: 1220,1010 (6.7mJy) and 2145,947 (3.4mJy) plus a few others. Let try some unconstrained cleaning. > imagename='testmosaic2.clean' > niter=50 > mosaic Worked. Ideally, if I run this again I get 50 ADDITIONAL iterations. Lets check this out. First copy testmosaic2.clean.* ==> testmosaic2.clean.n50.* Then, > mosaic Yes, it clearly starts from the previous state - excellent! Could now pbcorrect by dividing by fluxscale using ia tool, but am testing tasks. NOTE: would be good to have pbcorrect tool This worked well. What the heck, lets image the whole mosaic (106 fields): 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 Note last row is incomplete in this day's data. Center should be field 62. Size should be something like [10000,2500] which will be like 300MB. > imagename='testmosaic4.dirty' > niter=0 > imsize=[10000,2500] > fieldid=[62]+range(2,62)+range(63,106) > phasecenterid=62 > saveinputs 'mosaic','mosaic4.saved' > mosaic terminate called after throwing an instance of 'std::bad_alloc' what(): St9bad_alloc Abort (core dumped) Must have tried for too much memory (but this was on basho)! Kumar suggests was due to it being non-square. NOTE: core dumps when making huge non-square image. Leaves around some big TempLattice files too! NOTE: I tried to run qtviewer while mosaic is running and got the read lock error Done testing.