Re: loadStringN can write to freed memory, will fail assert in tests

[Roberto Ierusalimschy <roberto@...>]

> I am afraid that change does not ensure that there is a GC while reading
> the long string, so it may not trigger the bug. (It didn't when I tried.)
> The following chunk seems more reliable:
> 
> ---------------------------
> -- run this chunck under some memory checker
> local function myload (s)
>   return load(function ()
>   [...]
> ---------------------------

I think this function is already implemented in testes/call.lua under
the name 'read1'.

-- Roberto