On 8 December 2015 at 22:53, Stefan Dösinger stefandoesinger@gmail.com wrote:
However, the reason why I've never made this seemingly obvious change is that you'll get other applications that crash due to VM exhaustion now. So I'm afraid you'll just trade one kind of breakage for another. The proper fix is probably somewhere in GL_ARB_buffer_storage, but that is too complicated for 1.8.
Yeah, there's that risk. It seems preferable though, and there are probably more options for reducing vm usage / fragmentation than for fixing this bug. I suppose I'm also hoping that applications that need a lot of vm space know better than to use the managed pool, but perhaps that's too optimistic.
That said, I'm not particularly attached to fixing this bug before 1.8, and the application in question needs additional patches to even start. The patch could certainly be deferred if you think the risk isn't worth it.