The portage ebuild compiles the sandbox with really low CFLAGS. It should just use the regular ones in make.conf. Reproducible: Always Steps to Reproduce:
possibly because the sandbox developers (az and the portage team) know better than to trust user CFLAGS for something as critical as the sandbox? Just a guess.
I thought of something like that, but if you're going to set crazy CFLAGS, you might as well be aware of the consequences. I don't know how much really low CFLAGS are going to slow down portage; my guess is not much, considering the sandbox is only used for a little part of the emerge process, but I'm really looking for ways to optimize, and I think that compiling the sandbox without user CFLAGS is pretty unnecessary.
And the sandbox isn't all THAT critical. How many ebuilds really violate the sandbox? The only time its critical at all is if a user is screwing around with an overlay.
Except that if your sandbox lib is broken, you can't compile anything...
if the sandbox segfaults, it'll pretty much take out any and all processess running sandbox'd. Meanwhile, marking this later. the autotooling of sandbox allows for it, although the question of filtering/limiting cflags hasn't been addressed.
Reopening for reassignment
this has already been addressed
