Re: [Avr-libc-corelib] Error handling

[Ruud Vlaming]

On Monday 21 September 2009 23:27, Jan Waclawek wrote:

> A common thing to do is to return an error flag if a function fails to 
> perform as intended.....
Although common, i usually find it a waste of the costly return variable, 
which, as you
correcty point out is usually ignored. (If that is wise is an other matter ... )

> Ergo, IMHO it's a pure waste to return runtime error codes, and I suggest to 
> establish a single variable, 
> core_lib_error or such, where functions could store an appropriate code when 
> error occurs, and 
> leave it upon the user how does he wish to use or ignore it.   
Allthough i agree that mostly error codes are a waste (if they do not provide 
specific information)
i strongly oppose the use of a single error variable. This makes concurrent use 
of the library
virtually impossible. Lets make all code reentrant, or at least up to the point 
where
the hardware dictates otherwise.

But i think we don't have 'errors' in the classical way. An 'error' is a 
situation that the called
function cannot handle. But i would say, that should not happen, or at least, 
since our
state space of the functions is usually very small, the 'error' should be a 
natural part 
of the returned information, thus, no generic error handling. 

Ruud.