On Mon, Jul 27, 2009 at 8:37 AM, Charles Oliver Nutter<[email protected]> wrote: > On Sun, Jul 26, 2009 at 12:15 PM, Yehuda Katz<[email protected]> wrote: >> There are two different issues here: >> 1) how many frames up a backref goes back to >> 2) whether it's possible for a method to modify ITS backref but still have a >> previous backref (in a given thread) still valid in some context >> I don't think (2) is actually possible, and (1) is solved by a single >> thread-local slot, which eliminates the need to care about propagating back >> up. > > I think Yehuda is probably right. We often have had to go up more than > one frame (like for gsub with a block, for example) but we only have > ever had to set it in one place. So I think we may still be ok. It > will be worth prototyping, in any case.
Yeah certainly worth a try. At the very least we can jot down why it doesn't work if there is a problem. For a moment I was thinking it was block_given? but I think that is ok. -Tom -- blog: http://blog.enebo.com twitter: tom_enebo mail: [email protected] --------------------------------------------------------------------- To unsubscribe from this list, please visit: http://xircles.codehaus.org/manage_email
