Re: "*toclose" vs. sandboxes / embedded environments

[Andrew Gierth <andrew@...>]

>>>>> "Roberto" == Roberto Ierusalimschy <roberto@inf.puc-rio.br> writes:

 Roberto> Wouldn't be enough some internal flag to disable all
 Roberto> operations? An error would set that flag, and some explicit
 Roberto> operation (done after the pcall, in case of errors) would
 Roberto> reset it?

 >> Yes, it's an option I'm considering. The main downside is that it
 >> would have to be checked in a lot of places, and there are places in
 >> the code (such as in __gc metamethods) where it's not enough just to
 >> disable operations.

 Roberto> Another option: can't you do whatever is being done in the end
 Roberto> of 'pcall' when you raise the error?

The error is being raised somewhere outside of my code, I don't have any
control over that. I do catch and rethrow it, but then we're back to the
problem of always having to create subtransactions.

Hmm... what is the relative timing between *toclose cleanups and the
error handler for xpcall? that wasn't immediately apparent from the code.

-- 
Andrew.