GNOME Bugzilla – Bug 125436
Gnumeric crashes after attempting `linest' function
Last modified: 2004-12-22 21:47:04 UTC
error in line 518 of regression.c (general_linear_regression): should not be reached. Gnumeric crashes after attempting `linest' function on 60 data points. y-data and x-data seperated by 10 orders of magnitude (1e10). Steps to reproduce the crash: 1. Get data for a line 2. Separate x- and y- data by at least 10 orders of magnitude (e.g x/1e10) 3. =linest(x1:x10,y10:y10) Expected Results: Finds estimate of gradient. How often does this happen? Only when x- and y- values separated by 10 or more orders of magnitude. Possibly problem with algorithm? - If this is so, perhaps just a check to make sure the data is valid (i.e. not separated by 1e10), and to not do calculations if so rather than crash. Additional Information: running gnumeric-1.2.1-bonobo on linux 2.4.20 under blackbox 0.65.
Created attachment 20924 [details] bug report autogenerated when gnumeric crashed
Works for me. Please supply _precise_ instructions on how you generate this. (What values to put in what cells, etc.)
Got it. Pondering fix.
Fixed in cvs. (This is a loss-of-precision case so you will end up getting #NUM errors.)