Re: Fwd: NSUserName() returns wrong user and..

[David Ayers]

Richard Frith-Macdonald wrote:

> Any thoughts on this?
>
> Does anyone object to changing NSUserName() to return the name
> associated with geteuid() rather than the LOGNAME environment
> variable?

I see no legitmate reason to use an environment variable to to control 
the current user.  In fact 'su user' (without the -) also changes the 
whoami output and cd changes to the new users home directory.

> Should the name be (as currently) cached, or should the function
> check the effective user id every time it is called?

I think caching should be left up to the application.  If it asks 
mulitple times, there could be a good reason for it.

> At present there is an extra function GSSetUserName() which can
> be used to manually set the username and reset the user defaults
> system ... but if NSUserName() rechecked geteuid() every time it
> was called, it could reset the user defaults system itsself.

Actually, I see no inherent connection between querying the user name 
and updating the user defaults.  I'd rather ask if this function should 
actually /do/ the setting of the user (i.e. call setuid) rather than 
merely syncronize with the current user id.

To support 'external' setuid maybe we could extract the syncronization 
in a GSSyncronizeUser() or similar function.  OTOH we could discourage 
that and say that we provide GSSetUserName() for that, and then, I 
guess, we could again cache the user name for NSUserName().

Cheers,
David

PS: We might want to call the funtion to set the user GSSetUser() and 
deprecate GSSetUserName().