Hi Michel,
I am not at work today, but I think I know what is happening. I remember I noted that on octave gsa crashes if I do not type explicitly
clear global;
before running it after a previous dynare session.
There is a check in dynare_sensitivity for the existence of estim_params_ which triggers or not dynare_estimation_init. Without clear global there is still a variable estim_init_ in the octave workspace that skips estimaton init and therefore the routine crashes. I will have to check if this check is really necessary and perhaps put an explicit initialization flag there.
Still, this clear global issue seems to be something to be aware of and fix for octave sessions?
best Marco
On Sun, 30 Oct 2011 18:08:42 +0100, Michel Juillard michel.juillard@mjui.fr wrote:
Super! We have greatly diminished the number of failures on the tests. I tried to see why gsa/ls2003.mod was failing under Octave but can't understand. The test works when dynare is called interactively, but fails when called via run_test_octave.m. It looks as if the interaction of try and global estim_params_ is creating a problem in this case. It would be great if somebody finds out what is going on.
Best
Michel
Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev