[MPlayer-dev-eng] [PATCH] libmpdvdkit2/dvd_reader.c: do not second-guess user

[Rich Felker]

On Thu, Oct 20, 2005 at 10:55:55AM +0200, Arpi wrote:
> but libdvdread also recognizes if the given path is mounted from a device
> (based on mount tab), and then uses the raw device itself instead of the
> filesystem. it's required to handle mounted css-encrypted disks.
> (before you ask: why dont the user give the device insteda of the path
> in this case? because on some OSes it's quite difficult to know the device
> name, esp. on device mappers and other messy devfs-like systems, while you

IMO this is not true. The device name is stored in /etc/fstab and is
quite constant, and a properly setup system will always have a symlink
to it (named something like /dev/dvd).

> can easily know the mount point, not even mentioning windows, where you
> dont have device names, just drives which are actually paths...)

This is why I want windows to be handled specially.. Other such
systems could be too..

> the best would be adding a -dvd-filesystem or such option, and patch
> libdvdread to handle it.
> so if you use -dvd-device path, it will know you want to use the device,
> located based on the given path, and if you use -dvd-filesystem  then
> you want to use file access instead of raw sector-based addressing.

I suppose this is ok, but I still don't like libdvdread doing scanning
for the device name. Anyway the current code is INCORRECT, since it
assumes getmntent exists, which is NOT required by POSIX/SUS.
getmntent is legacy bsd crap, and I've removed it from my (in
progress) libc replacement because it's not useful for any program
except mount/umount/df (basic system progs which can include their own
code, or better yet use /proc)..

Rich