Received: (at submit) by bugs.debian.org; 20 Nov 1999 15:04:42 +0000 Received: (qmail 9004 invoked from network); 20 Nov 1999 15:04:42 -0000 Received: from finch-post-12.mail.demon.net (194.217.242.41) by master.debian.org with SMTP; 20 Nov 1999 15:04:42 -0000 Received: from watchdragon.demon.co.uk ([158.152.121.201]) by finch-post-12.mail.demon.net with esmtp (Exim 2.12 #1) id 11pC2n-00004t-0C for submit@bugs.debian.org; Sat, 20 Nov 1999 15:03:17 +0000 Received: from localhost (mnementh.local) [127.0.0.1] by watchdragon.demon.co.uk with esmtp (Exim 3.03 #1) id 11p0FA-0003Uk-00 (Debian); Sat, 20 Nov 1999 02:27:16 +0000 To: submit@bugs.debian.org From: pmaydell@chiark.greenend.org.uk Subject: cruft: uses too much disk space Date: Sat, 20 Nov 1999 02:27:16 +0000 Sender: pm215@watchdragon.demon.co.uk Message-Id: Package: cruft Version: 0.9.5 Severity: wishlist The way cruft is designed causes it to use rather a lot of disk space. For example, when producing a report on my fairly small system (filesystems 1GB, 250MB, 250MB) it put 15MB of stuff into /var/spool/cruft/. Surely this could be reduced somewhat by a cleverer algorithm? For example, /usr/lib/cruft/explain/dev is just a 'find /dev' command. This means that all the files in /dev are listed in a file in the spool. It would be better to allow something like a wildcard or regexp syntax so you didn't have to list all the files in /dev explicitly. This gets much worse if you use this strategy for directories like /usr/local, naturally. I think that a better way to do the job would be to first create the files to be used to filter the file names, as we do now (expl_* and need_*), but then to do the weeding out of 'OK' files on the fly as part of the pipeline 'find $DRIVE...', rather than creating large file_* files and then processing them. Obviously this would be quite a bit of work :-> Peter Maydell