[PATCH] do not call _dbus_warn_check_failed on checks

Havoc Pennington hp at redhat.com
Tue Nov 14 10:52:30 PST 2006


Kevin Krammer wrote:
> 
> I think this is even documented somewhere: libdbus is not supposed to be used 
> directly but rather through some binding.

Yep, this is documented on the front page of the manual.

> Summing it up there is a need for plain-C bindings, similar to the plain-C++ 
> bindings someone else wrote about a couple of weeks ago.

I don't think nice plain C bindings would do what Daniel is advocating 
though; Daniel's suggestion IMO makes the API even more low-level, not 
higher-level. To go higher-level a binding needs to make more 
assumptions and have even more default behaviors.

The main point of me arguing at all is so everyone else is clear on the 
what's and why's; if someone else is doing a binding (or just using 
libdbus, or hacking on libdbus), they should understand the semantics 
and guarantees of the current API, and how to write correct code using it.

That means understanding the difference between a bug/programming-error 
and a recoverable error, and that libdbus (like the GLib/GTK+ stack and 
I believe the Qt stack also which last I checked did not return an error 
code from every method) makes a distinction between how the two are handled.

Havoc



More information about the dbus mailing list