|
From: Leonard N. <leo...@re...> - 2003-07-22 01:20:09
|
Hi all, A small step on the way to write-support: I just added Richard Russons initial documentation on the B*Tree operations on the NTFS directory structure, which he wrote last year, to the web site. See the B*Trees section at http://linux-ntfs.sourceforge.net/ntfs/concepts/index.html - don't miss the references section, or the illustrations of the add and delete examples. This is of interest to people wanting to work on the NTFS write code, probably beginning with adding support for renaming files. That step is a a big amount of work, though. The initial code for B*Tree support was written last year by Richard Russon. It's available by installing BitKeeper and then issuing the command: bk clone bk://linux-ntfs.bkbits.net/tng-support The tree code lives in tng-support/tree for now, but is nothing more than rudimentary tree manipulation code as a standalone program and a shell script "random" to feed it random keys/nodes. For now, it only uses a single character as the node/key name. It does work, however (or so says flatcap :-). It needs some cleanup and conditional compiling on the debug code so it can be merged into a kernel tree. I say we need to add automated unit testing to keep write support related code bugfree. Some unit test frameworks are available here: http://www.xprogramming.com/software.htm and you should read up on automated test frameworks here http://c2.com/cgi/wiki?TestingFramework in case the concept is new to you. People wanting to work on the write code would be wise to get UML, User-mode Linux (see http://user-mode-linux.sf.net/ and if you're using Linux 2.6: Documentation/uml/UserModeLinux-HOWTO.txt). UML has a nice feature that lets you work with a master disk image file, while directing any changes to a separate file - copy on write. Thus you can experiment without fear of data loss. For more information on how to do this, see http://user-mode-linux.sourceforge.net/UserModeLinux-HOWTO-7.html Have fun, Leonard Norrgard |
