Damian Conway via RT <perlbug-followup@perl.org> wrote: > I have now had the opportunity to investigate the problem, and have concluded that this has nothing to do with Regexp::Grammars per se, except that R::G is generating the enormous regex that 5.20 is failing to compile. Yes, it looks that way to me too. Thanks for supplying that reduction. The file attached cuts down this regex further still, by removing all the embedded code blocks, and the various DEFINEs whose names begin "______0_88". The symptoms I observed seem to be the same, though I also get a "panic: memory wrap" error (apparently when passing 4 GiB of allocated memory). In blead, it looks like the immediate culprit is study_chunk() — it starts 185 ms after the start of the sizing pass, and I haven't yet had the patience to let it run long enough to finish. -- Aaron Crane ** http://aaroncrane.co.uk/Thread Previous | Thread Next