On Wed, 21 Dec 2011, Tim Jenness wrote:
> >> I wonder if this is a provenance thing although provenance should
> >> only be active if an input file has provenance and none of your steps
> >> will enable provenance (unless $AUTOPROV environment variable is
> >> set).
> >
> >
> > I checked and this variable is not set.
> >
> > Now that I've trying other commands I have noticed that this "thinking"
> > state seems to occur also when using findobj and tranndf. Again, if run over
> > a small number of images the commands produce their output very quickly but
> > if I do the same over a large number of images the extra time used by the
> > routines is not proportional to the additional number of images (and by
> > checking the contents of the folder it seems that the routines have already
> > done their job when the verbose output ends).
> >
> >
>
> There are two ways forward:
>
> 1. You run findobj in the debugger and ctrl-c it during its "pause"
> phase and tell me what routine it is stuck in
> 2. You tar up some test data and put it somewhere so I can run a
> test (tell me the command you use). Don't send the test data to the
> whole list.
No need, I can easily reproduce the problem by just feeding a set of fake
images into CALCOR (made by repeated calls to KAPPA:CREFRAME). It does
look like a provenance issue as that's where debugging leads me to and if
I do:
setenv AUTOPROV 0
and then run CALCOR on 1000 images it returns immediately. Eduardo this
will be a good fix for you.
Here's the backtrace for David to ponder:
#0 0x00002afb417c69da in ndg1ProvCmp (a=0x2a6b4a0, b=0x2a6b898) at ndg_provenance.c:6061
#1 0x00002afb45ad2744 in msort_with_tmp (p=0x7fff5c9b1ca0, b=0x2a6b3c8, n=234) at msort.c:84
#2 0x00002afb45ad2521 in msort_with_tmp (n=234, b=0x2a6b3c8, p=0x7fff5c9b1ca0) at msort.c:46
#3 msort_with_tmp (p=0x7fff5c9b1ca0, b=0x2a6b3c8, n=469) at msort.c:54
#4 0x00002afb45ad2521 in msort_with_tmp (n=469, b=0x2a6b3c8, p=0x7fff5c9b1ca0) at msort.c:46
#5 msort_with_tmp (p=0x7fff5c9b1ca0, b=0x2a6b3c8, n=938) at msort.c:54
#6 0x00002afb45ad2a3b in msort_with_tmp (p=0x7fff5c9b1ca0, n=938, b=0x2a6b3c8) at msort.c:46
#7 qsort_r (b=0x2a6b3c8, n=938, s=<optimized out>, cmp=0x2afb417c69bc <ndg1ProvCmp>, arg=0x0) at msort.c:298
#8 0x00002afb417c6b87 in ndg1PurgeProvenance (provenance=0x2a884a8, status=0x7fff5c9b2bbc) at ndg_provenance.c:6145
#9 0x00002afb417bf03a in ndgPutProv (prov=0x5f517, indf=192225282, more=0x0, more2=0x0, isroot=0, status=0x7fff5c9b2bbc) at ndg_provenance.c:2515
#10 0x00002afb417bd227 in ndg_putprov_ (iprov=0x7fff5c9b27c0, indf=0x7fff5c9b27c4, fmore=0x2afb417dc628 "<NOT A LOCATOR> HistorySKIP", more2=0x2afb417dc624, isroot=0x2afb417dc5d8, status=0x7fff5c9b2bbc, fmore_length=16) at ndg_provenance.c:1094
#11 0x00002afb417d6dd4 in ndg_endpv (creatr=Cannot access memory at address 0x7fff5c9b2830) at ndg_endpv.f:326
#12 0x0000000000405d4c in ccdpack_red (status=0) at ccdpack_red.f:134
#13 0x0000000000405af1 in dtask_applic (context=<optimized out>, actcode=<optimized out>, aname=<optimized out>, actptr=<optimized out>, seq=0, value=Cannot access memory at address 0x7fff5c9b2970) at dtask_applic.f:75
#14 0x00002afb44cf0ffb in dtask_obeydcl (dtask_applic=0x405a40 <dtask_applic>, name=<optimized out>, value=..., status=0, _name=<optimized out>, _value=444) at dts_obeydcl.f:161
#15 0x00002afb44cef7c7 in dtask_dcltask (devinit=0x405860 <devinit>, dtask_applic=0x405a40 <dtask_applic>, status=0) at dts_dcltask.f:153
#16 0x0000000000405a2f in maintask () at dtask_main.f:150
#17 main (argc=<optimized out>, argv=<optimized out>) at dtask_main.f:172
#18 0x00002afb45ab930d in __libc_start_main (main=0x405870 <main>, argc=1, ubp_av=0x7fff5c9b2ca8, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fff5c9b2c98) at libc-start.c:226
#19 0x0000000000405799 in _start ()
Note that none of the input images has a provenance.
Cheers,
Peter.
----
Starlink User Support list
For list configuration, including subscribing to and unsubscribing from the list, see
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=STARLINK
|