Re: [patch] enable unload-module

>>>>> "Sean" == Sean MacLennan <seanm(a)storm.ca&gt; writes: Hrvoje> Then why unload the module in the first place? Sean> You need this whilst writing the modules. Anyone who can handle writing a module can handle make CFLAGS='-DDANGEROUS_SCARY_MONSTER'. to make a module-testing platform. We're not talking about removing the code. Sean> I would not even care if there where *very* strict rules Sean> about what you where allowed/not allowed to do with a Sean> dynamic module if you wanted to unload it. But there's nothing we can enforce. I'm not even sure that Andrew's proposal for immediate reinitialization is good enough; in fact, I'm pretty sure it isn't. If you change and recompile the module before the unload, reload, reinitialize cycle, there's no reason to suppose that the subrs and symbols would have the same addresses. That means that any copies of those values taken before the reload would be dangling references. No, that is _not_ perverse or limited to module developers who can handle it, either. Suppose the module comes in an RPM? In order to detect that kind of change we'd have to use a unique-id setup like the Linux kernel does with modversions. So basically to be unloadable, the module would have to have no Lisp-visible features. ;-)