Re: Turn error on opening file using Process and TSelector in Fatal

From: Constantin Loizides <loizides_at_MIT.EDU>
Date: Wed, 17 Feb 2010 15:33:27 +0100


Hi Rene,
yes one can do this of course, but it
does not guarantee that when "at analysis time" (x mins laters) the file can be opened. Constantin

On 02/17/2010 03:34 PM, Rene Brun wrote:
> Hi Constantin,
>
> Why don't you make a test when adding the file to the chain? in this
> case, you can take the appropriate action yourself and not include the
> file in the chain if the file does not exist, or timeout or any of the
> zillion possibilities that you can imagine.
>
> Rene Brun
>
> Constantin Loizides wrote:
>> Dear all,
>> is it possible to intercept failing open commands
>> (in a TChain) when using a TSelector?
>>
>> In many cases the standard behavior of ROOT to skip and
>> go on to the next file is ok, but there are cases where
>> you really want to abort your analysis when the file
>> could not be opened? (In the case I am talking about
>> it is a file on castor that simply timed-out and was
>> later accessible again, so opening all files beforehand
>> is not fool proof).
>>
>> To me, the only (easy) way seems to overload the default
>> error handler or set the gErrorAbortLevel = kError.
>> But this would then treat all errors as fatal.
>>
>> Any ideas?
>>
>> Thanks,
>> Constantin
>>
>
Received on Wed Feb 17 2010 - 15:33:37 CET

This archive was generated by hypermail 2.2.0 : Wed Feb 17 2010 - 17:50:01 CET