Opened 7 years ago
Last modified 6 years ago
#1084 assigned defect
Reproducible thread error in bumps on Mac OS
Reported by: | smk78 | Owned by: | wojciech |
---|---|---|---|
Priority: | blocker | Milestone: | SasView 4.3.0 |
Component: | SasView | Keywords: | |
Cc: | Work Package: | SasView Bug Fixing |
Description
Reported by user Judith Witte (using Mac OS 10.13.3).
Was fitting happily in an (unspecified) earlier version of SasView which then just "stopped working" (wouldn't open anymore). So upgraded to 4.1.2. Data loads in 4.1.2 but on attempting a fit this is returned:
Error(s) Occurred: > Traceback (most recent call last): > File "sas/sasgui/perspectives/fitting/fit_thread.pyc", line 82, in compute > File "sas/sasgui/perspectives/fitting/fit_thread.pyc", line 14, in map_apply > File "sas/sasgui/perspectives/fitting/fit_thread.pyc", line 11, in map_getattr > File "sas/sascalc/fit/BumpsFitting.pyc", line 275, in fit > File "sas/sascalc/fit/BumpsFitting.pyc", line 375, in run_bumps > AttributeError: 'FitDriver' object has no attribute 'fitter'
This looks remarkably similar to #1061 which is only classified as minor, however Judith says this is stopping her from doing any fitting. She also comments that "The first error occurs after I choose the model (not clicked on Fit yet)". If you look carefully at the screenshots you will note two slightly different errors. So am initially marking this as a blocker pending further investigation as it appears a more serious and widespread problem than in 1061.
@pkienzle has commented:
"There is an error in bumps, but it isn't being properly reported because of an error in the bumps wrapper. I can arrange it so that the next bumps release fixes the wrapper error but I can't address the bumps error without being able to reproduce it."
Data and screenshots from Judith to assist diagnosis are attached.
Attachments (5)
Change History (11)
Changed 7 years ago by smk78
comment:1 Changed 7 years ago by pkienzle
XCode command line tools not installed and OpenCL turned off. Options are:
(1) default to opencl on for Mac (we shut it off because Intel HD6000 drivers were failing for some models; Intel has since fixed their drivers, though they may or may not have been distributed to all macs yet).
(2) force XCode command line tools to install when SasView first starts.
(3) ship tinycc for mac
comment:2 Changed 6 years ago by butler
As per discussion at SasView call of April 17, 2018 it is agreed that:
- OpenCL should be left off by default even though the original problem we think has been fixed. Because the GPU calculations are sensitive to how they are coded for the card, it is always possible that in a future upgrade to cards or the OpenCL stack that our code could start failing "silently." Users should use the test function built into SasView before turning on OpenCL.
- The right answer is the third option of shipping tinycc for mac. There is a start at this and it should not be too difficult but probably cannot be done for this release
Thus to resolve this ticket for release 4.2 we will just improve the error reported so that it is more obvious why it failed and create a new ticket #1095 for release 4.3 named "Ship tinycc with mac"
comment:3 Changed 6 years ago by butler
- Owner changed from pkienzle to wojciech
- Status changed from new to assigned
comment:4 Changed 6 years ago by pkienzle
Additional option:
(4) precompile models and include them with the application. Set the error message to give instructions for installing XCode command line tools in case they try creating their own C model or download one from the marketplace.
This option has the advantage that most users won't see it, or at least not until they are already familiar with SasView.
comment:5 Changed 6 years ago by butler
- Milestone changed from SasView 4.2.0 to SasView 4.3.0
For release 4.2 Wojcieck has implemented a more helpful warning to the user of why things are failing. Leaving this ticket open but moving to 4.3 as it is not clear that it has been resolved that we won't try to do something more active (i.e. some of the other options above.) If not we can close the ticket.
Data from Judith