[Gretl-devel] Re: latest Windows snapshot failing
On Wed, Nov 6, 2024 at 8:23 AM Cottrell, Allin <cottrell(a)wfu.edu> wrote:
On Wed, Nov 6, 2024 at 6:31 AM Sven Schreiber
<sven.schreiber(a)fu-berlin.de> wrote:
>
> the sample script of my package PTconf (current version 1.2) runs fine
> on 2024c and on a snapshot from Oct 25th, but is failing with an error
> on yesterday's snapshot. (error with new starting obs)
>
> Maybe the recent work on subsampling didn't quite get there yet?
The failure is just on t1, t2 (non-complex) sampling. I didn't think
my recent changes had touched that point but evidently they did,
somehow. I'm looking into it.
OK, this was to do with the --no-missing variant of "smpl": it was
failing to respect a prior restriction on the sample via t1 and t2.
Specifically, in PTconf: the sample script does "smpl 1960:1 2000:4",
then setupPT() does "smpl --no-missing endo" (where @endo has no
missing values). The second smpl command then overrode the incoming
restriction, erroneously restoring the full dataset range.
This is now fixed in git.
Allin