tophat - long_spanning_reads problem

[fabferre]

Hi there,
while running tophat, a file was created (left_kept_reads.fq.candidate_hits.sam) by the long_spanning_reads binary that is larger than 400G and that clogged the disc.
I'm using the hg18 assembly, and 17 million single end reads in fasta format. I tried both tophat-1.0.12 and tophat-1.0.13, using the default parameters, and both bowtie-0.11.3 and bowtie-0.12.1. I understand that tophat creates several very large temporary files, but I hope that >400G corresponds to some kind of error.
The following is the log until I stopped the process:

[Tue Jun 1 14:40:59 2010] Beginning TopHat run (v1.0.13)
-----------------------------------------------
[Tue Jun 1 14:40:59 2010] Preparing output location liver_out/
[Tue Jun 1 14:40:59 2010] Checking for Bowtie index files
[Tue Jun 1 14:40:59 2010] Checking for reference FASTA file
[Tue Jun 1 14:41:00 2010] Checking for Bowtie
Bowtie version: 0.12.1.0
[Tue Jun 1 14:41:01 2010] Checking reads
seed length: 32bp
format: fasta
[Tue Jun 1 14:42:17 2010] Mapping reads against hg18 with Bowtie
[Tue Jun 1 14:58:20 2010] Joining segment hits
[Tue Jun 1 14:58:56 2010] Searching for junctions via segment mapping
[Tue Jun 1 15:04:17 2010] Retrieving sequences for splices
[Tue Jun 1 15:04:17 2010] Indexing splices
[Tue Jun 1 15:04:18 2010] Mapping reads against segment_juncs with Bowtie
[Tue Jun 1 15:09:28 2010] Joining segment hits

Cheers