1 Revision history for Eval-Closure
4 - support lexical subs on 5.18+
7 - Remove a double layer of string eval that was introduced in 0.07 as an
8 intermediate step in figuring out the unique package thing - it's not
9 necessary with the final implementation, and just makes things slower
13 - the given source is now evaled in a unique package for every
14 eval_closure call (it used to always be evaled in the Eval::Closure
15 package, which was especially buggy). this is to avoid issues where one
16 eval_closure modifies the global environment (by, say, importing a
17 function), which could mess up a later call. unfortunately, this means
18 that the memoization stuff no longer works, since it will result in
19 memoized results using the original package, which defeats the purpose.
20 i'm open to suggestions on how to safely reenable it though.
22 - clean up a few stray lexicals we were still closing over in the eval
25 - The dumped source from setting $ENV{EVAL_CLOSURE_PRINT_SOURCE} didn't
26 show the captured variable creation properly. (Dave Rolsky)
29 - make sure Perl::Tidy doesn't look at @ARGV (rafl)
32 - stop using Memoize, it apparently doesn't work properly under mod_perl
33 in some situations (mateu)
36 - don't add #line directives when the debugger is active (alh)
39 - add terse_error option for throwing only the actual compilation error
41 - allow also specifying the line number for error reporting