I have just uploaded the tarball for shed skin 0.8. please test it and
report any problems, before I announce things further.
it has been a wonderful cycle for me, with feedback from several directions.
thank you all for that!
here's the (full) list of changes:
my personal highlights:
- the c64 emulator example was synced with its author, and now weighs in at
about 3,500 lines (sloccount). it still compiles in about 25 mins here,
which is in line with the scalability graph I published before. see
attachment for a screenshot of one of my all-time favorite games running on
it. using cpython, it runs at around 3 FPS, and after compilation, it's
perfectly fluent (at least 50 FPS).
- this triggered me to finally add support for the array and struct modules!
struct is used to read programs from tape, and array to contain a packed
version of the screen data that is passed out of the extension module.
- three new examples were added, most interestingly to me a fast (havlak)
loop recognition algorithm. please keep those examples coming.. ;) I'm
planning on doing a type inference round for 0.9, and I can't have enough
test cases for that..
- a new option, shedskin -x, makes generated code prints tracebacks for all
thrown exceptions. so it also print them when they are actually caught.
still I think it can be very useful to investigate why a large program
crashes (such as a c64 emulator). I would be interested if anyone knows how
to improve this so we only print tracebacks for uncaught exceptions (that
is, efficiently of course).
- I just added a 'faq' section to the wiki section (
http://code.google.com/p/shedskin/wiki/faq) with responses to some common
questions/preconceptions about the approach shedskin takes.
thanks again to those that helped with or provided feedback for this
if there are no issues reported with the current tarball, I will proceed to
package the windows version and ask thomas and paul to upgrade the rpm/deb