Re: 'Suspect' libglib-2.0 when launching Ardour2.8.4

On Dec 15, 2009, at 5:59 PM, Paul Davis wrote:

> it was still a remarkable screw up on their part - compiling glib for
> x86 as if it was a platform with no atomic operations, and thus
> forcing all "atomic" ops to actually use mutexes, thus totally
> changing their semantics. i question whether glib should even have
> this fallback - it totally changes the semantics of a program if
> g_atomic_int_add() can block, for example ....

That's actually a pretty understandable goof. A system where uname reports 'i386' will, by default, not build the atomic operations. Since most macs do that, Gtk-OSX is configured to tell glib "--build=i486-apple-darwin" to force building the atomic asm functions. 

It might be reasonable to get rid of that quirk in glib's configure, as I rather doubt that many (more like any) folks are using actual 386 boxes these days.

John Ralls

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]