Re: cdrecord floating point exception

[Matthias Andree <matthias.andree@gmx.de>]

Joerg Schilling schrieb:
> Bill Davidsen <davidsen@tmr.com> wrote:
>
>   
>>> Try to learn that hald on Linux is broken and acts on wrong status changes.
>>>   
>>>       
>> Nothing is ever your fault. Instead of learning from the applications 
>> which burn CDs and DVDs without being root, your software has problems 
>> with hald and you refuse to accept that changing the hald config fixes 
>> the problems and others can work with hald as is, and insist hald is at 
>> fault.
>>     
>
> besides the fact that you _need_ root privileges in order to do the tasks 
> cdrecord does,
Not true.

1. You tend to mix up the goal and the way to get there. Your goals are:
being permitted to open a device, send raw commands, lock pages into
memory, and use real-time scheduling, perhaps more. There is not only
one way there. Your way there is to require super user permissions,
which means that at the same time you get permission for all privileged
operations at the same time.
It goes without saying that from a programmer's POV super user
permission is an easily solution, since that's universal (not
system-specific) thus you need not port this part to different operating
systems.

2. You mix up cdrecord's ambitions and those other applications may
have. Other applications may decide to restrict their offerings when
they cannot obtain all privileges. For instance, an application that
cannot obtain realtime permissions and cannot lock pages into memory may
refuse to write media that do not allow single-block overwrites. Such
applications can still write DVD-whatever. Or, if the user requests
reduced quality and enables Burnproof/Justlink/whatever, can still write
a CD without realtime scheduling.

HOWEVER your wording often reads as though the decisions you made for
cdrecord were the only possible ones. That's true for cdrecord and for
software that you write, but not for other applications, or other
programmers.
> I am of course willing to help the hald people to fix their software.
>   
After a bit of behind-the-scenes discussion with Thomas of scdbackup
fame, I start wondering if Linux's device access model is up to the task.

Let's collect some facts first, before we start pointing fingers at anyone.

How do operating systems (Solaris/openSolaris, *BSD, Linux) provide
EXCLUSIVE access to a device?
I'd naively tend to believe that if MyCdWriteApplication has
/dev/blahwriter0 open for exclusive access, no other application should
be able to bypass that exclusive reservation. Yet Thomas claims he's
seen such things happen. I'm not sure if that's a user-space or
kernel-space problem.