On Apr 6, 2012, at 4:28 PM, Joerg Sonnenberger wrote: > On Fri, Apr 06, 2012 at 06:21:06PM -0000, John McCall wrote: >> Log: >> Use atexit when __cxa_atexit isn't available instead of adding a >> global destructor entry. For some reason this isn't enabled for >> apple-kexts; it'd be good to have documentation for that. > > Are you sure that this won't run into atexit(3) limits? Some platforms > IIRC had a somewhat tight limit for those.
This is the traditional behavior of -fno-use-cxa-atexit as documented by GCC, at least for static locals. We're very likely using atexit in places where we could and should be using a global destructor instead, but those are separable bugs. I mean, personally, I think this has been standardized for over a decade and people ought to have started shipping __cxa_atexit on their platforms by now, but if they still haven't gotten their act together, their users would probably rather build and run into implementation limits than not be able to build at all. John. _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
