On Thu, 28 May 2015, Sven Schreiber wrote:
Am 20.05.2015 um 18:58 schrieb Sven Schreiber:
> Am 20.05.2015 um 16:16 schrieb Allin Cottrell:
>> On Tue, 19 May 2015, Allin Cottrell wrote:
>>
>>> On Tue, 19 May 2015, Sven Schreiber wrote:
>>>
>>>> I'm doing a VECM simulation with restricted estimates. Adding the
>>>> options "--full --silent" after the "end restrict"
already does 99%
>>>> of what I want, namely to not produce printed output. However, two or
>>>> three lines of messages with respect to degrees of freedom of the
>>>> Jacobian remain.
> still seeing on stderr (or what's equivalent to stderr on Windows, I
> mean in the command prompt window where I launch gretlcli.exe) is the
> message "warning: test df = 0", or actually thousands of them...
> I tried "set warnings off", but I think that only relates to stdout not
> to stderr and in any case it didn't work. So is it possible to suppress
> these warnings?
The more I think about it, I think that this type of warning message
should also be suppressable by "set warnings off". Currently the doc of
"set warnings" says it only concerns "messages issued when arithmetical
operations produce non-finite values", but why limit it to that?
In any case, on the compute cluster where I have run these gretl
simulations I'm getting error log files of 20MB each just because of
thousands of these messages... AFAICS I cannot work around this with
'outfile null --write' because this does not affect stderr.
The usual idiom to discard whatever appears on stderr is
gretlcli -b input.inp 2>/dev/null
Allin