Every time I run Picard's AddOrReplaceReadGroups, my system tmp directory fills, and the tool stops. I'm left scratching my head, because it seems that the JVM sets the system /tmp directory on my local galaxy-central branch, instead of the "new_file_path" I have defined in universe_wsgi.ini. This issue seems to have been fixed in commit 07560293b167, and I can confirm that my picard_wrapper.py includes the amended line. I am using OpenJDK:

java version "1.6.0_20"

OpenJDK Runtime Environment (IcedTea6 1.9.10) (rhel-1.40.1.9.10.el6_1-x86_64)

OpenJDK 64-Bit Server VM (build 19.0-b09, mixed mode)

Since my config is pointing to /scratch/tmp, but the JVM is using /tmp, and other tools in galaxy are respecting the config, what could be the disconnect? Thank you,

Matt