I did so. Unfortunately on AIX I cannot test now (cern afs cell not
responding..) But on SGI INDY I tested it and the problem with reading
back the events remains. In an interactive root session I can read
and print run & event of the first 2836 events before a seg violation.
Every event contains around 40 TClonesArrays (no other containers). The
file has 100 Mb (10000 events).
It's not a sick event: if I start reading at event 2836 it crashes at
event 5672 thus I think there is a buffer overflow somewhere.
The problem is absent in batch. Here I can read all w/o any problem.
Both the batch job and root-session had a const memory allocation during
the reading of events.
machine: SGI INDY single processor 256 Mb RAM, IRIX 6.2, SCSI disk or
NFS-mount (I tested both)
root: version 1.03/08 25 November 1997
ahoi
Jobst
On Tue, 9 Dec 1997, Rene Brun wrote:
> Hi Jobst,
> Couild you change you buffer size from 64K to say 16K.
> I see that you have a very large number of branches. Root needs twice
> the space
> for memory buffers (to compress) at the end of the job when you write
> the Tree
> header on the file.
> Rene
>
>