El sáb, 12-12-2009 a las 18:34 +0100, Sven Schreiber escribió: I repeated your steps firstly with gretl built on december 9 (on Ubuntu Linux 9.10) and after that with gretl built just now, and in both cases I don't obtain a crash. The output is: Sistema VECM, orden del retardo 4 estimaciones de Máxima Verosimilitud, observaciones 1973:1-1991:1 (T = 73) Rango de cointegración = 1 Caso 3: constante no restringida beta (vectores cointegrantes, Desviaciones típicas entre paréntesis) PAU 1,0000 (0,0000) PUS -3,0156 (0,51320) IAU_1 -2,5207 (9,2520) IAU_2 15,363 (8,3597) IUS_1 11,141 (7,9130) IUS_2 -5,1118 (7,3574) alpha (alfa (vectores de ajuste)) PAU 0,0034811 PUS -0,0096879 Log-verosimilitud = -142,69443 Determinante de la matriz de covarianzas = 0,17096136 AIC = 4,4026 BIC = 4,9674 HQC = 4,6277 Ecuación 1: d_PAU Coeficiente Desv. Típica Estadístico t Valor p --------------------------------------------------------------- const 0,723281 0,527800 1,370 0,1754 d_PAU_1 0,139629 0,140211 0,9958 0,3231 d_PAU_2 0,315540 0,146587 2,153 0,0352 ** d_PAU_3 0,120829 0,144805 0,8344 0,4072 d_PUS_1 0,562600 0,245402 2,293 0,0252 ** d_PUS_2 −0,440802 0,264517 −1,666 0,1006 d_PUS_3 −0,0805612 0,252506 −0,3190 0,7507 EC1 0,00348110 0,00590694 0,5893 0,5578 Media de la vble. dep. 2,376712 D.T. de la vble. dep. 1,135478 Suma de cuad. residuos 66,11593 D.T. de la regresión 1,024431 R-cuadrado 0,287777 R-cuadrado corregido 0,186031 rho −0,018969 Durbin-Watson 1,922460 Ecuación 2: d_PUS Coeficiente Desv. Típica Estadístico t Valor p --------------------------------------------------------------- const 0,878983 0,246449 3,567 0,0007 *** d_PAU_1 −0,0341883 0,0654699 −0,5222 0,6034 d_PAU_2 0,0967910 0,0684471 1,414 0,1623 d_PAU_3 −0,0153595 0,0676148 −0,2272 0,8210 d_PUS_1 0,228302 0,114587 1,992 0,0507 * d_PUS_2 0,120999 0,123513 0,9796 0,3310 d_PUS_3 0,426381 0,117904 3,616 0,0006 *** EC1 −0,00968793 0,00275817 −3,512 0,0008 *** Media de la vble. dep. 1,178082 D.T. de la vble. dep. 0,628987 Suma de cuad. residuos 14,41529 D.T. de la regresión 0,478345 R-cuadrado 0,493933 R-cuadrado corregido 0,421637 rho 0,042825 Durbin-Watson 1,890975 Matriz de covarianzas cruzadas entre ecuaciones: PAU PUS PAU 0,90570 -0,088806 PUS -0,088806 0,19747 determinante = 0,170961 -- Ignacio Diaz-Emparanza DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA) UPV/EHU Avda. Lehendakari Aguirre, 83 | 48015 BILBAO T.: +34 946013732 | F.: +34 946013754 www.ea3.ehu.es

Sven Schreiber schrieb: well strangely enough it's now even worse for me: the crash already occurs at the first estimation stage, without modifying the spec. It also does when running gretl in English. The console error output is simply "segmentation fault". I will test on Windows tomorrow. cheers, sven

On Mon, 14 Dec 2009, Sven Schreiber wrote: Unless you updated from CVS in the last 2 minutes, the problem was not solved! Sometimes you "get lucky" and a memory management fault doesn't cause a crash. But in fact this is unlucky from the coder's point of view because it masks the problem. Hint: anyone using gretl built from CVS might want to consider building with debugging symbols on ("-g") and then running gretl under valgrind when this sort of issue occurs. Valgrind usually manages to spot exactly where the error is, whether or not it actually produces a crash. Allin

Allin Cottrell schrieb: ah ok, interesting that the bug was kind of erratic. So are you saying that it will necessarily show up under valgrind if "it needs more work"? If yes, I could try to do that. Or is the issue more subtle? thanks, sven