On Thu, 4 Apr 2019, Sven Schreiber wrote:
Am 03.04.2019 um 23:32 schrieb Allin Cottrell:
> On Wed, 3 Apr 2019, Sven Schreiber wrote:
>> Personally I'm getting errors that I didn't see on another system.
I'm
>> not sure if my configuration on the new system is somehow wrong or
>> whether there is a deeper issue, possibly some incompatibility with
>> Windows 10 (of the package itself or of the Tramo program).
It cannot be Windows 10 per se, since I have tested on another Win10
system where it works fine.
> Thanks, as ever, for maintaining a rigorous testing standard! I just
> tried the sample script of version 1.11 (2017-01-28) of the tramolin
> package on Linux and it ran OK. But I guess that's not very relevant;
> I'll have a go on Windows 10 tomorrow.
I've now tested on Windows 10 and the sample script ran OK.
Thanks Allin. It's not clear to me what triggers the error on
one
machine but not on the other. But I noticed that in both cases an empty
text file 'sername.txt' was produced in the respective current work
directory. From Ignacio's code I didn't really understand why it ended
up in those locations, it looked as if it was intended to go into
$tramodir somewhere. But since on both machines the file is created and
is empty, this doesn't seem to be the (sole) reason.
I see where the empty sername.txt is coming from (it appears on Linux
as well as Windows): it's produced by this (shell command) line in the
tramolin function:
! cat "@tramodir/ausentesf" >> sername.txt
The write location for sername.txt on this line does not carry over
from the previous line, namely,
! cd @seriesdir ; <write stuff to sername.txt in @seriesdir>
since it's a new invocation of the shell. And (at least in the case of
the sample script) the file @tramodir/ausentesf is empty. If
"ausentesf" should really be appended (if not empty), the line that
handles it should read something like:
! cat "@tramodir/ausentesf" >> @seriesdir/sername.txt
I don't know if that's actually wanted. (Note that sername.txt in its
location under the @dotdir/tramo tree is the file that contains the
linearized series to be added to the gretl dataset via the "append"
command.)
The only thing that caught my attention was that in the relevant
directory paths on the failing machine there was a hyphen character (-).
I guess this _shouldn't_ cause a problem but who knows.
Seems unlikely. But in which path(s) does the hyphen appear? Does it
occur in @tramodir?
For troubleshooting, you might try taking a look at
@dotdir/tramo/graph/series/sername.txt
I guess that's the file that's provoking the "Invalid data file"
message -- maybe it's empty or corrupted? (If it were missing
altogether I think you'd get a different message.) Also, does
@dotdir/tramo/graph/series/xlin.t
exist, and if so what does it look like?
Allin