On Sun, 5 May 2013, Marcin Błażejowski wrote:
When I've tried to generate 'fcast --out-of-sample
--no-stats --quiet'
based on arima model I got crash and following GDB results:
---
Program received signal SIGSEGV, Segmentation fault.
create_Xb_series (Z=0xfd8460, xlist=0xf87d50, beta=0x0, pmod=0xee47c0,
fc=<optimized out>)
at ../lib/src/forecast.c:1273
1273 x = arima_difference_obs(Z[vi], t, delta, k);
---
Note, that it happens whene I generate forecasts inside the function
package. In normal script it works fine:
---
open bjg
smpl 1949:01 1959:12
arima 0 1 1 ; 0 1 1 ; lg --quiet
fcast --out-of-sample --no-stats
---
Can you give some more context, please? I tried the following:
<hansl>
function void fcbug (series y)
arima 0 1 1 ; 0 1 1 ; y --quiet
fcast --out-of-sample --no-stats
end function
open bjg
smpl 1949:01 1959:12
fcbug(lg)
</hansl>
This runs OK and gets a clean bill of health from valgrind.
I also tried moving the smpl command inside the function;
still no problem. So it seems we need more background to
replicate the bug.
Allin