From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 5.0) Description of problem: Some data written into tmp does not get cleaned up and slowly fills the entire file system: either /tmp or / root if /tmp is part of it. The unremoved data is evident via "df" report but not visible via "ls" of "du" (thus appears to be raw disk data). I've only been able to cleanup the data via a destructive "mkfs". This problem is observed on all (about 10) of my systems with or without all the latest 7.1 Errata patches installed including 2.4.3 kernel patches. My installations include most all Redhat packages. The problem is visible even immediately following a fresh install after initial boot. Examples of programs that create the disk leakage are "gnu make" and the "df" command itself (sometimes df needs to be run several times for the /tmp or / disk space use to be registered ... less frequently the larger the fstab) ... not a corresponding increase with du -s /tmp command is observed. Interestingly this behavior (at least with df stimuli does not occur if rlogin'ed into the machine). It also seems to be partly dependent on being in the KDE env (df related leak does not seem to cccur when in a non-gui console window although the gnu make leak seems to still be present). Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1.enter kde; note size of /tmp with "du" and "df" 2.run numerous "df" commands (iterations needed to illustrate problem vary based on df table size) 3.see df of /tmp space increase while no increase in "du -s /tmp" or any new files/dirs visible under /tmp via "ls". Actual Results: /tmp or / disk space usage increased Expected Results: no increase in disk space usage Additional info: space not recoverable unless mkfs is done
What you describing is just a bug of konsole which has an unlimited history which fills the file system. (Just have a look at http://bugs.kde.org/db/30/3004.html ) The difference from df and du is just the size of the history of konsole.
True - and the fileutils tools are actually doing the right thing (by not counting changes in files that are still opened for writing). I agree that konsole history size should be limited though.
This is fixed in KDE 2.2, which was shipped with 7.2.