New GC Core
julian.notfound at gmail.com
Mon Jul 13 11:35:37 UTC 2009
> 2) This core does not EVER call any destroy VTABLEs, so things that
> rely on that (such as automatic flushing/closing of filehandles)
> break. The Parrot build mostly completes, although JSON notably relies
> on this behavior of FileHandles so it does not build
I fixed TGE pir generation by explicitly closing, and checking close
result, when the output is not stdout. I've done this to improve error
detection, not just for the benefit of this GC core, but the result is
that now parrot build completely and pass a lot more tests.
I've even be able to build rakudo and pass his tests. It even started
to pass spectest, but I killed it because at some point it started to
eat all memory on my system (not surprisingly, beacuse it not so close
to infinity as desired).
More information about the parrot-dev