g_assert vs. g_return[_val]_if_fail

[Reinhard Mueller <reinhard mueller bytewise at>]

Hi all,

We are using glib all over our project (GNU Enterprise) and want to make 
best use of it's debugging functions.

Now the discussion has started on in which cases g_assert should be used 
and in which g_return... is better.

We know the differences:
* g_assert is turned off by a different define than g_return..., but 
both can be turned off.
* g_assert terminates the application, where g_return... only terminates 
the function
* therefore, g_assert lets me jump into gdb as soon as it is hit, while 
g_return... doesn not. This seems to be an advantage for g_assert.
* for users, it seems that g_return... is nicer because it doesn't kill 
the process.

So my questions are:
Is there a golden rule when to use which?
Does it make sense to use both and disable the g_assert for production 
use and let g_return... in?
How do others handle this?

Thanks a lot,
--
Reinhard Mueller
GNU Enterprise project
http://www.gnue.org