[sigc] remarks on using (or not using) sigc++
- From: Paul Davis <paul linuxaudiosystems com>
- To: "Libsigc++" <libsigc-list gnome org>
- Subject: [sigc] remarks on using (or not using) sigc++
- Date: Tue, 16 Sep 2014 13:03:55 -0400
[ this is just a remark i sent privately to kjell, and he suggested i forward it to the whole list ]
the ardour project moved away from sigc++ because it is
not thread safe. we initially settled on boost::signals(2?) but then
found out that this not correctly thread safe (i don't recall the
details, but it was either not actually thread safe despite claims to
be, or insufficiently close to realtime safety for us to use)
we
ended up developing our own signal system which combines thread safety,
relatively good realtime safety and most of the basic features of
sigc++. we generally use boost::bind to construct the equivalent of
sigc::slot, which provides most of the fancier features :)
we
continue to use sigc++ for things that only involve the GUI, since
thread safety is not required there (it is mandatorily single-threaded
code).
we generate our signals.h header using python.
[
Date Prev][Date Next] [
Thread Prev][Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]