Opened 9 months ago

Closed 9 months ago

#875 closed defect (fixed)

Possible weirdness with 1D NXcanSAS data

Reported by: smk78 Owned by: butler
Priority: major Milestone: SasView 4.2.0
Component: SasView Keywords:
Cc: Work Package: SasView Bug Fixing

Description

During use of 4.0.1 in our training course this morning several students generated the attached error whilst fitting an ellipsoid * HayterMSA to the SANS from a 0.5% SDS in D2O solution.

There were two things in common: (i) they were all using identical lightweight Dell desktops (ie, I guess it could be a CPU/GPU issue???), and (ii) the data they were loading had been stored in the NXcanSAS .h5 format (actual file attached).

When I asked them to repeat their actions but using the equivalent data from the SasView test folder they could not generate the error.

I have been unable to replicate the error on my office desktop (where SAS_OPENCL=none).

Anyone any ideas what caused this error?

Attachments (2)

WP_20170308_001.jpg (483.5 KB) - added by smk78 9 months ago.
Error message
0.5 wt% SDS.h5 (48.0 KB) - added by smk78 9 months ago.
NXcanSAS file

Download all attachments as: .zip

Change History (4)

Changed 9 months ago by smk78

Error message

Changed 9 months ago by smk78

NXcanSAS file

comment:1 Changed 9 months ago by butler

  • Owner set to butler
  • Status changed from new to accepted

comment:2 Changed 9 months ago by butler

  • Resolution set to fixed
  • Status changed from accepted to closed

This is just the usual response from fitting that goes to a bad place - LM is particularly prone to this. However in this case the values used to start were clearly bad: the charge was negative and the temperature set to > 6100 K! The default min was set to 0 at some point prior to 4.1 and the default max temp has been set to something sensible: 450 K.

Will close this ticket but see about reopening ticket 820 or creating another one to capture the fact that a more useful pop up would be nice.

Note: See TracTickets for help on using tickets.