Rockbox mail archiveSubject: Re: FAT Filename Mangling bug. (was: RE : Regarding the recent recording PANIC's)
Re: FAT Filename Mangling bug. (was: RE : Regarding the recent recording PANIC's)
From: Björn Stenberg <bjorn_at_haxx.se>
Date: Fri, 25 Apr 2003 10:16:15 +0200
Holger Lembke wrote:
> > Only if you can convince me that I am actually violating the
> > specification and not just doing it differently than the win98 code.
> http://www.hlembke.de/lfnspec.pdf 3.2.2 and 3.2.4
Nice document. However, I get the feeling this too is a "how we will support long filenames in windows". Maybe it's just me.
My main point for doubting that the shortnames matter, however, is that there is no technical need for the shortname to match the longname. They are not connected by the name, but by three separate mechanisms each being much more secure than filename association:
1) Position in dir. The longname directory entries are _always_ positioned immediately prior to their associated shortname entry.
2) Checksum. Each longdir entry contains a checksum of the shortdir entry, to verify that they are indeed connected.
3) Numbering. Each dir entry in a long+shortname is numbered in a decreasing sequence, with the shortname being 0.
If any of these three rules are broken, the dir entry is invalid.
-- BjörnReceived on 2003-04-25